d2.2 successful scenarios for ontology-based applications ...q the clustering of applications in...

Deliverable 2.2 Final - 1 -

IST Project IST-2000-29243 OntoWeb

OntoWeb: Ontology-based Information Exchange for Knowledge Management and Electronic Commerce

D2.2 Successful Scenarios for Ontology-based Applications V1.0

OntoWebOntology-based information exchange for knowledge

management and electronic commerceIST-2000-29243

Date: 31st May 2002

Identifier Deliverable 2.2

Class Deliverable

Version 1.0

Version date 31-05-2002

Status Final Draft

Distribution Public

Responsible Partner France Telecom R&D

Deliverable 2.2 Final - 2 -

Project ref. no. IST-2000-29243

Project acronym OntoWeb

Project full title Ontology-based information exchange for knowledge managementand electronic commerce

Security (distribution level) Public deliverable

Contractual date of delivery 31 May 2002

Actual date of delivery 31 May 2002

Deliverable number D2.2

Deliverable nameSuccessful scenarios for ontology-based applications

Type Report

Status & versionFinal Version V1.0 – May 31st.

Number of pages 100 Pages

WP contributing to thedeliverable

WP2

WP / Task responsible WP2

Main contributors Alain Léger – France Télécom R&D, Yannick Bouillon, PhilippeEcoublet –Atlantide, Martin Bryan –the SGML center, Rose Dieng– INRIA Sophia, Andreas Persidis -Biovista, York Sure – Universityof Karlsruhe, Asuncion Gomez-Perez & Mariano Fernández López -UPM, Ying Ding - VUA

Editor(s) Alain Léger –France Télécom R&D

EC Project Officer Hans-George Stock

Keywords Ontology-based applications, Successful Business Scenarios, BestPractices, Guidelines for implementers, Key findings

Abstract (for dissemination)D2.2 focuses on the identification of the most suitable techniquesthat may be applied to each cluster of ontology-based applications.Its aim is to provide guidelines to assist an organisation on whattechniques could be applied for a given application.

OntoWeb. D2.2. Best Practices and Guidelines – V1.0 IST-2001-29243

3/100

Document History

Version Date Comments and Actions Status

V 0.1 25/03/2001 First draft version edited by co-ordinator thatintegrates WP2 contributions

Draft proposal

V 0.2 15/04/2001 Second draft version edited by co-ordinator thatintegrates WP2 contributions

Draft proposal

V 0.2.5 27/04/2002 Second enhance draft version edited by co-ordinator that integrates WP2 contributions andcomments

Draft proposal

V 0.3 30/04/2002Third draft version co-edited with contributions

Draft proposal

V 1.0 31/05/2002Final version Delivered to Project

Co-ordinator

AcknowledgementsThe Editor would like to thank specifically the following main contributors:

D2.1:ÿ� York Sure, Rudi Studer and SiegfriedHandschuh(University of Karlsruhe, Germany) for the

Knowledge Management part and the Semantic portal part;ÿ� Andreas Persidis(Biovista) for the Information Retrieval part and especially for the continuous

support of this work;ÿ� Rose Diengand her team (INRIA Sophia Antipolis, France) for the Knowledge Management

part;ÿ� Martin Bryan (The SGML Center, UK) for the E-Commerce part;ÿ� Mike Brown andAsuncion Gomez-Perezfor the global effort to edit D12 (“Business

Scenarios”) in harmony with D21 (that deliverable: “Successful Scenarios and Guidelines”);ÿ� Ying Ding for the constant effort to provide many materials for the completion of the D21;ÿ� Hans Akkermans for the information provided for the E-Commerce section and the global

effort to help in providing right resources within the SIG4 (Applications)ÿ� Jean Marc Bouladoux(Odos Strategy sa Paris) for contributions on Evaluation ;ÿ� Important Contributions on KBS Evaluation (so rare) based onTim Menzies papers and

referencesÿ� Contribution from Best Practices LLC (Susan Silverstein, Sales Executive)ÿ� Many others contributors (always referred in the References section) to theE-work and E-

Commerce 2001event held in Venice (15-17 October 2001) and sponsored by the EuropeanInitiative in New Methods of Work and Electronic Commerce, also known as Key Action IIin the European Information Society Technologies (IST) programme.

D2.2:Special thanks already toMartin Bryan (SGML Center), Rose Dieng (INRIA Sophia), AndreasPersidis (Biovista) andYork Sure (University of Karlsruhue) for their valuable and prompt contributionto D2.2, particularly on Knowledge Management, Information retrieval, Portal and web community.Particular acknowledgements also to the continuous support received fromAsuncion Gomez-Perez andMariano Fernández López(UPM), andYing Ding (VUA).Thank you for all other remarks and comments received from all other WP2 partners.


4/100

Executive summary

This deliverable is the secondof a series of 5 documentsto be delivered on a6 monthly basis andwhose target is to reportBest Practices and to give Guidelines for the application of Knowledge ITtopractitioners in the field of - in large - E-Commerce and Knowledge Management.The D2.2 is based onthe existing D2.1 deliverable.Its goals is to enhance the D2.1 content making more mature someidentified parts or sections and providing some new results. Moreover and in consistency with the originalcontract, D2.2 will be aimed at:

q Identification of the most suitable techniques that may be applied to each cluster of ontology-based applications.

q Provision of guidelines to assist an organisation onwhat techniques could be applied for a givencluster of applications.

The relationship between deliverables D 1.2 and D 2.2 is as follows:q Deliverable D 1.2 (Business Scenarios) distils from this information and other sources the

reasonsWhy and underWhat conditions a commercially viable ontology application can bedeployed. Deliverable D 1.2 also offers some predictions as to future commercial trends.

q Deliverable D2.2 (Best Practices and Guidelines) goes further in attempting to provide aconcrete set of guidelinesas to what needs to be achievedin order to deploy ontologysolutions, i.e. this deliverable describes theHow to construct assuccessfulontology-basedapplication.

This series of documentsis primarily aimed at anyone who is involved in the process of designing,building and managing Knowledge-Based systemsfor the field mentioned above. Itshould assistanyone in the industry or the commercial sectors to evaluate how these emerging technologies aresuitable to answer practical needs. Furthermore, each successful retained scenario is illustrated bypractical examples.

Considering the deliverable D2.1 as a background for the deliverable D2.2 is aiming at:

q Enhancing the parts dedicated to applications and business scenarios, particularly those aiming atB2B, B2C, KM and IR.

q Justifying the use of ontologies to enhanceprofitability both from organisational orfinancialaspects, inconsistencywith already identified business scenarios,

q Identifying usability criteria’s (generic or specific) for ontology-based applications and providingsome evaluation guidelines for usability, in consistency with the previous point,

q Building a structured evaluation framework for ontology-based applications to:o Selectkey or killer applications ,o Identify bothgeneric or specific criteria that can be used in each cluster of applications,o Provide anevaluation grid as an evaluation framework.

q Applying the evaluation framework to a set of key application examples trying to get as outputs anestimation of benefits and defining somekey characteristics for business model.

q Providing some recommendations to verify, demonstrate and evaluate the technical and economicsuitability of ontology-based systems or applications, in consistency with some selected businessscenarios.


5/100

Our Key Findings from the D2.2 study and editing work, should be summarized as follows:

q Globally the evaluation istruly favouring the use of Ontology-based systemsin the class ofapplications fully covered in that report (IR and Enterprise portals / KM);

q Many parts of the Evaluation and Guidelines study (D2.2) have now reached agood level ofmaturity : "Overall evaluation framework", "most representative applications", "Evaluation and"Guidelines in IR and Enterprise portals / KM";

q The Information Retrieval class of services seems well covered in every facets of the study carriedout in D2.2, probably due to the relative maturity of the subject. The D2.2 givesclear answerson IREvaluation and Guidelines.

q The Knowledge Managementclass of services seems also rather well covered. The D2.2 givesrather clear (even a bit less formalized than IR) answers on KM Evaluation and Guidelines.

q The E-Commerce and Community portalshave appeareddifficult to evaluate. We were not ableto locate public references on(semi-) formal evaluation of those classes of applications. Thatevaluation exists – at least for the E-Commerce sector – butthe most pertinent ones are notpublicly available. However, we believe that the good coverage of the IR and KM classes could helpgreatly in filling the gap in the other two main classes of applications. This is a clear planned studyfor the D23 deliverable.In spite of this D2.2 have found some good directions for the guidelinessection.

q The clustering of applications in main classes like E-Commerce B2B and B2C, Enterprise portals andKnowledge Management, Information Retrieval and Portals for communities, seemed difficult andsomewhat arbitrary, and this fact remotes from the start of writing the D2.x series. So, trying toabstract the situation led us to consider that a"web services" orientation of the thinking could helpin the future series of deliverables.


6/100

Table of contentsExecutive summary___________________________________________________________________4

1 SCOPE___________________________________________________________________________9

1.1 Introduction ___________________________________________________________________9

1.2 Business Models for the Semantic Web (?) _________________________________________10

1.3 Terminology and Acronyms _____________________________________________________12

2 Why evaluate Ontology based applications?_____________________________________________13

2.1 Motivation ___________________________________________________________________13

2.2 Basis for an Evaluation Framework of Ontology based Applications ___________________142.2.1 Major classical objectives of an evaluation _______________________________________142.2.2 Measuring technology using financial measurements _______________________________162.2.3 Human Assessment__________________________________________________________17

2.3 References ___________________________________________________________________18

3 Most Representative Applications _____________________________________________________19

3.1 Introduction __________________________________________________________________19

3.2 Enterprise Portals and Knowledge Management____________________________________213.2.1 Service definition ___________________________________________________________223.2.2 Use Case and Needs _________________________________________________________233.2.3 Business-to-Employee (B2E) as a most Representative Applications ___________________243.2.4 References ________________________________________________________________25

3.3 E-Commerce _________________________________________________________________263.3.1 Why are ontologies promising for e-Commerce?___________________________________263.3.2 Service definition ___________________________________________________________263.3.3 Use Cases and Needs in B2C __________________________________________________273.3.4 Use Cases and Needs for B2B _________________________________________________283.3.5 Representative Applications using Ontologies for E-Commerce _______________________293.3.6 References ________________________________________________________________31

3.4 Information Retrieval __________________________________________________________323.4.1 Service definition ___________________________________________________________323.4.2 Use Cases and Needs ________________________________________________________333.4.3 Representative Applications ___________________________________________________343.4.4 References ________________________________________________________________38

3.5 Portals and Web communities ___________________________________________________393.5.1 Service definition ___________________________________________________________393.5.2 Use Cases and Needs ________________________________________________________403.5.3 Representative Applications ___________________________________________________413.5.4 References ________________________________________________________________44

4 Evaluation of Ontology Based Applications_____________________________________________45

4.1 Why evaluate ontology-based applications? ________________________________________45

4.2 Criteria or metrics for the evaluation of ontologies __________________________________464.2.1 Generic Criteria ____________________________________________________________474.2.2 Usability __________________________________________________________________484.2.3 Expressivity _______________________________________________________________484.2.4 Accuracy__________________________________________________________________494.2.5 Consistency________________________________________________________________49


7/100

4.2.6 Completeness ______________________________________________________________494.2.7 Conciseness _______________________________________________________________504.2.8 Expandability ______________________________________________________________504.2.9 Sensitiveness_______________________________________________________________50

4.3 Evaluation in Information Retrieval ______________________________________________514.3.1 Reference evaluation methodology _____________________________________________514.3.2 Reference evaluation criteria __________________________________________________524.3.3 Specific Ontology enhanced criteria_____________________________________________54

4.4 Enterprise Portals and Knowledge Management____________________________________574.4.1 Reference evaluation methodology _____________________________________________574.4.2 Reference evaluation criteria __________________________________________________574.4.3 Specific Ontology enhanced criteria_____________________________________________61

4.5 E-Commerce _________________________________________________________________624.5.1 Reference evaluation methodology _____________________________________________624.5.2 Reference evaluation criteria __________________________________________________624.5.3 Specific Ontology enhanced criteria_____________________________________________62

4.6 Portals and Web communities ___________________________________________________624.6.1 Reference evaluation methodology _____________________________________________624.6.2 Reference evaluation criteria __________________________________________________624.6.3 Specific Ontology enhanced criteria_____________________________________________62

4.7 References ___________________________________________________________________62

5 Successful Scenarios and Guidelines __________________________________________________63

5.1 Introduction __________________________________________________________________63

5.2 Corporate Intranet and Knowledge Management ___________________________________635.2.1 Guidelines_________________________________________________________________645.2.2 Keys to the success in building an enterprise portal_________________________________65

5.3 E-Commerce _________________________________________________________________675.3.1 Guidelines for E-Business: B2C and B2B ________________________________________67

5.4 Information Retrieval __________________________________________________________725.4.1 Guidelines_________________________________________________________________72

5.5 Portals and Web communities ___________________________________________________765.5.1 Guidelines_________________________________________________________________76

5.6 Other approaches to be used to develop domain ontologies ___________________________77

5.7 References in Guidelines________________________________________________________79

6 Tools and Methodology for Ontology-based Applications __________________________________81

6.1 Tools for ontology-based applications _____________________________________________81

6.2 Ontology engineering __________________________________________________________836.2.1 Methodology_______________________________________________________________836.2.2 Related Work ______________________________________________________________85

6.3 Ontology building from text _____________________________________________________86

6.4 References ___________________________________________________________________87

7 Conclusions ______________________________________________________________________88

8 References _______________________________________________________________________89


8/100

Annex 1 SEAL Description ___________________________________________________________93

8.1 History ______________________________________________________________________93

8.2 Web Information Integration____________________________________________________93

8.3 Web Site Management _________________________________________________________94

Annex 2 Creating a SEAL-based Web Site_______________________________________________95

Annex 3 Representative applications for corporate intranet and Knowledge Management_________97

8.4 OntoBroker __________________________________________________________________97

8.5 OntoKnowledge _______________________________________________________________97

8.6 CoMMA (Corporate Memory Management through Agents). _________________________99


9/100

1 SCOPE

1.1 IntroductionKnowledge-based technologyis increasingly being applied to a very large set of applications inside,which formalized knowledge and data, may be processed by computers. Thus, some of these applicationsor services will now mix human readable and structured data so that both humans and computers can usethem. To name the most prominent ones we can enumerate:Enterprise Portalsand KnowledgeManagement, E-Commerce, Natural Language processingand Machine Translation, InformationRetrieval, Integration of Heterogeneous Information, IntranetCommunity Portals, etc.

From a business viewpoint, such technology may really benefit to the entire society, first to the economy,allowing companies to better interoperate and quickly find new or best opportunities, and second, tocitizens or customers because it will help them to support their daily work, leisure and interactions withorganizations. As a consequence to this, interest in the evaluation of the Knowledge-based technologyis growing more and more due to theincreased maturity of the field and due tobusiness pressureforthe evaluation of themeasurable costs and benefits of that technology.

Usually, such evaluation is quite easy to manage since the main purpose of a company is to create valuethat customers are willing to pay for and then, to estimate the production, the market share and then thereturn on investment (ROI) for corresponding products or services. Within our study, the main difficultyis to evaluate thebusiness added valuea company may benefit if it makes the choice to use ontology-based technologies which are not directly bought by customers but which are aiming at helping peopleneeds i.e.service added value(e-commerce web sites, mobile employees connecting to their enterpriseportal, etc.).

Moreover, before trying to evaluate the added value ontology-based applications may create, it may bepreviously efficient to answer the following question: for what purpose are these technologies to be used?

Thus, beingahead of the competitioninvolves employingkey technologiesand whenever they existcapturing essential guidelines and key success factors, trying to keep in mind the aforementionedquestion. Unfortunately, as a matter of facts, quantitative studies on that field arerelatively rare, dueprobably to theyouth of this technologyand also – as a key competitive knowledge –kept private ineach business practices.

As previously mentioned, the major goal of the OntoWeb WP2 group is to issueBest Practices andGuidelines in providing benchmarking methodologies based on the performance characteristicsofvarious ontology-based systems designed for differentclasses of applications:

Each Best Practice and Guideline document series will describe the class of system or services beingaddressed; discuss the performance characteristics that are pertinent to that class; clearly identify a set ofmetrics that aid in the description of those characteristics; specify the methodologies required to collectsaid metrics; and lastly, present the requirements for the common unambiguous reporting of Best Practiceresults.

Benchmarking is the process of seeking out, studying and quantifying the best practices that producesuperior performance. The traditional metrics-focused approach must be supplement with an analysis ofwhy and how practices produce better results. It should helpunderstanding strengths and weaknessesand can be supplemented by a road map for improvement (companion document D1.2). Thedetermination of benchmarks allows one to make a direct comparison of off-the-shelves bundle solutionsor its own solution against competitors. Any identified gaps in the comparative study are potentialimprovements.


10/100

Another form of benchmarking includes"process benchmarking", generally higher-level and lessnumbers-intensive than metrics. These studies demonstrate how top-of-class application accomplishes thespecific process in question.

Best practices are documented strategies and process employed by top-of-class applications. One givenapplication can not claimed to be top-of-class in every area - such application does not exist - so in D2.xand series, we are capturing the best practices of the top best applications in a given class.

We have collected this information from a variety of sources available from the OntoWeb Network as awhole. This can take the form of interviews, experiences, surveys, publications, books, magazines,libraries and Internet.

1.2 Business Models for the Semantic Web (?)1

Here is given the kind of related questions that the study currently pursued in the D2.x series is trying toanswer.

ÿ� The Semantic Web, even business-wise, is only at its very beginnings. Many aspects are still unclear:

ÿ� What are the specific problems for which identifiable, reachable purchasers will buysolutions that are based on Semantic Web technology, and which cannot reasonably beimplemented any other way? As an example, consider eBay for today's web: eBay clearlyfills a need, and could not exist in any other medium. In other words, what makes a solutionbased on theSemantic Web "10 times" better than what we have today?

ÿ� What will be the emerging "business ecosystem" around the Semantic Web? What arereasonable product/solution categories? Some potential categories are: Ontology modelingtools, ontology servers, servers to execute logical assertions, equivalence/ontologytransformation servers, etc. Is the Semantic Web going to be a client-server system, or apeer-to-peer system with symmetric protocols? Is there going to be a stack of technologieson top of each other, or even a knowledge stack?

ÿ� Is the Semantic Web primarilyan opportunity for upstarts with disruptive technologies,or an opportunity for established companies to providemore value to their existingcustomer as part of regular product upgrades?

ÿ� How many of the technologies currently being developed under the banner of the SemanticWeb are "real" from a business perspective? How many more do we need to be able toimplement Whole Products for mainstream users?

ÿ� Will the compelling reason to buy Semantic Web technologies be primarily one ofincreased revenue or lowered cost for the customer? Or one that is non-monetary incase of consumer adoption?

ÿ� Advertising as a business model for the Semantic Web is clearly not going to work, as theSemantic Web is all about communication between computers, who don't care aboutadvertisements. What are the dominant business models going to be?

ÿ� Even more fundamentally: are we going to see adoption first in the business markets or inthe consumer markets? How are we going to solve thechicken-and-egg-problembetweeninformation/knowledge providers (who will not do so unless there are users) andinformation/knowledge users? (Who will not make investments in software, education,change of behavior unless there is a Semantic Web out there already)

ÿ� Who are going to bethe primary producers of Semantic Web content?The case can bemade that many existing information providers on the web have strong incentives againstproviding their content in machine-processable form, as that would lower customer

1Excerpt from International Semantic Web Working Symposium, Stanford University, CA, USA August 1, 2001


11/100

switching costs to competitors If it is not them, who else is it going to be, and how do theyassemble critical mass?

ÿ� Is the adoption of theSemantic Web going to occur because of:

ÿ� Social dynamics (one can argue that today's web initially was adopted not for businessreasons, but because it was "cool" in its initial user segment)

ÿ� The increased availability of XML-based data sourcesÿ� The need for enterprise, and inter-enterprise application integrationÿ� The supply of knowledge brokers/online brokers for businessÿ� Web servicesÿ� The need for machine-to-machine communication?

ÿ� There are manyhuman factors that need to be considered, such as:

ÿ� Why will people annotate their web content, in particular if they do not benefit directlythemselves?

ÿ� How many content authors are even qualified to use technologies such as ontologies? Howcould one grow that number? Or is our current technology just far too hard to use?

ÿ� What kinds of authoring tools are necessary to make this process simple? Do they have to bedomain-specific, and if so, how could they be built and distributed efficiently by softwarecompanies?

ÿ� How many people are really qualified to build ontologies? How could we foster ontologyreuse?

ÿ� How could the Semantic Web piggyback on existing work processes in order to reduce costof Semantic Web content creation?


12/100

1.3 Terminology and Acronyms

We list here some of the terms and acronyms used throughout this document that refer to ontology basedapplications:

Ontology application scenario: A scenario is an abstract use case for a class of similar applications.Application means a system or a process that makes use of or benefits from the ontology. It describes aparticular situation in which ontology is put to use for some specific purposes.

Actors: Each scenario involves a set of actors. Each actor represents a role that a person or applicationmay play. A person or application may play more than one role in a scenario.

Application User: the role of the user of the application

Benchmarking: Benchmarking is the process of seeking out, studying and quantifying the best practicesthat produce superior performance. The traditional metrics-focused approach must be supplement with ananalysis of why and how practices produce better results. It should help understanding strengths andweaknesses and can be supplement by a road map for improvement. The determination of benchmarksallows one to make a direct comparison. Any identified gaps are improvement areas.

Metrics: It gives numerical standards against which a client’s own processes can be compared. Metricbenchmarks are of the form:

ÿ� Precision and Recall is better than 90% for Ontology-based matchingÿ� Ontology maintenance Cycle time is less than 2 weeks

These metrics are usually determined via a detailed and carefully analyzed survey or interviews.Applications developers are then able to identify shortcomings, prioritize action items, and then conductfollow-on studies to determine methods of improvement.

ROI: It gives the Return On Investment


13/100

2 Why evaluate Ontology based applications?

2.1 Motivation

In the increasingly competitive knowledge-intensive economy, the search forcompetitive advantageandthe necessity to optimally allocate scarce resources creates a pressing need forevaluating the ROI, aswell as thecontribution to value creation of applications. This is especially true of IT and knowledge-intensive applications for which this contribution is not always clear and often difficult to justify.

The semantic web is a relatively new concept whose stated vision is generally accepted by industry, eventhough there are as yetvery few successful high-profile casesthat support the theory and demonstratetangible results. However, in order to avoid mistakes that have been made in the past with other leadingedge, high profile technologies such as AI, it is essential that the semantic web and the ontologycommunities devise evaluation methodologies that they will be able to apply from the initial stages oftheir effort. There are two main reasons why this is useful:

ÿ� Management of expectations: As with all new things it is always wise to “promise less and delivermore”. Evaluation of ontology-based systems will indicate the scope and limitations of suchapplications thus allowing developers to understand the capabilities of the technology anddeployment time frames and end users to be educated about system potential.

ÿ� Guidance of the development and implementation process: evaluation of applications is useful todevelopers since it provides information and feedback that can be used in their further development.

How are weto assessall the Knowledge Engineering Technology being reported in the Semantic Webliterature or advertised on the web by popping-up very young companies? How can wemake themtransferable to the Industrial and Commercial community? And above all, how can we make itbeneficial for and transparent to the final user as any today ordinary service. We should carefullyassess superlative claims usually made about new technologies and especially in the very burgeoning areaof Semantic Web technology.

For example in the KA literature, the currently available empirical evidence does not support many of theclaims in the PSM literature. [Menzies, 1997], [Menzies, 1998]

Clearly, we need somebetter method than trusting the glowing reports available from the promotersof these Knowledge technology or embedded solutions. Even if these authors are experts in their fields,they are very often unable to perform objective expert evaluations. Cohen compares this situation to aparent gushing over the achievements of their children and comments that...

What we need is not opinions or impressions, but relatively objective measures ofperformance.[Cohen, 1995]

Taking into consideration the wide spread of the potential use of the Knowledge Engineering Technologyin applications or solutions,the task of defining objective criteria and methodology to assessSemantic Web Solutions seems extremely challenging.


14/100

2.2 Basis for an Evaluation Framework of Ontology based Applications

In deliverable D2.1 we have reported some well established - whenever known - criteria and methodologywithin each cluster of semantic web applications. In that first document, we clearly made a distinctionbetween:

ÿ� Internal properties of a Ontology-basedsystem (e.g. syntactic anomalies: tautologies, cycles) asusually performed by the verification community,

ÿ� Service and Business level evaluation(e.g. utility, return-on-investment, economic suitability) onlyaddressing the levels of services and business.

Few authors have tried to face the problem of service and business levels evaluation, and sovery fewreported proposals and practical conclusions is available.The main objectives of this second deliverable D2.2 is now to progress on that matter, in order to outlinethe generic properties of an evaluation framework for Service and Business assessment.

2.2.1 Major classical objectives of an evaluation

The major classical objectives of the evaluation are:

ÿ� To verify, to demonstrate and to evaluate the technical and economic suitability of the system tocommercial deployed applications,

ÿ� To evaluate the ontology-based application results and to provide recommendations of refinementsand enhancements

ÿ� To define rough characteristics of business models and fields for which they could be useful.

Figure 2.1- Facing Evaluation of Ontology-Based Systems

ÿþýüýûýúùø÷öÿ�û÷�ý��û�

ÿ��öø�÷ùø�úÿ��÷øúÿ�




ÿ��øúý��ÿ��ù�û

ÿ��ÿ� �� ÿ��ÿ� � �

ÿ�úü��öý�ù��ÿ�þ��ÿ� � �

ÿ�� ÿ��þ�� ÿ� � �

ÿ�÷ú�ö��ÿ� ø�û��ÿ� ��ýý��ÿ� � �

ÿþýüýþûúù

��ý��ú��þ��úù

��ý��ú��þ��úù

��úù

ÿþýüýþûúù

��ý��ú��þ��úù

��ý��ú��þ��úù

��úùÿ��øúý��ÿ��ù�û

ÿ��øúý��ÿ��ù�û


15/100

One aspect of the evaluation in such emerging technology is also to carefully precise the real innovationof the "new" technology, application or service, even different from that assumed by the proposal orinitially sought:

What was the state of the art improved in? It may lie in several places: for instance, in a newordering of existing functionalities, bringing thereby new or stronger services (as for robustness /relevance / quickness / ease of use / …) ; or in the technical field, opening possibilities of radicallynew services ; or in the strength of the gathered system developer team, eager to develop suitablesolutions to some of their questions, and willing or not to offer them on the market ; … Most of thetime, it is of course a mix of these, the evaluation purpose being then to split the different levels andprecise each of them,

What is this innovation worth? What are its worth and the worth of its uses in its supposeddifferent application fields (they may be threefold: breakthrough, reinforcing or economical), and thevalue proposition it can support? How long could it give a differentiation advantage to its users andowners? What efforts and means would be necessary to fix and feed this advantage? Is there asufficient plinth for so doing, as for available skills, technical documentation, …

What are the potential competitors, whether existing or to emerge?In particular, what othertechnical means, possibly simpler or more efficient, may emerge within some time? How follow theiremerging?

What marketing schemes seem the most relevant, throughout months: first targeted fields, firstenvisaged restricted uses of the product, kinds of partnership to seek? What particular businessmodels could be derived from these schemes?

The point is to “file down” the system results,to reduce the system to its very gist, out of all itsclothing, essential though when it comes to judge its usability (evaluating the product in a peculiar use,that of the retained business cases, while perceiving all its generic interest); this cannot be done by thecontender team on its own, even if it is the only one able to lead this evaluation.

The team in charge of evaluation purpose is as a consultancy: commissioned by the project leader, it muststudy round the product, make up its mind its own way through finding support even technical outside thecontender team, andstand up for its point of view in front of its client ; this latter remainsthe solejudge of the relevance of the conclusions and master of their publication – without twisting them.

The different pieces of advice, opinions, and feelings will usually be gathered:

That of each developer (or developing team):among other questions, what are they proud of, howdo they express the value proposition of the developed product, what did they learn, what originalknowing did they summon up, what improvement could be brought, what are the remainingdifficulties, what were the initial objectives modified in, who masters the whole system?

That of each test user: among other points, what is the perceived innovation and what is itssupposed importance (worth of its uses), in what field should it be the most immediately interesting,what business model appear to be the most obvious as for them?

That of the first potential clients: the commercial or marketing teams of the partners associated tothe development are questioned on the same topics than test users, plus improvements andrefinements necessary before adding the product into their existing commercial offer; if allowed bythe project achievement, it will also include testing of packaged pre-commercial proposals (includingfirst version of commercial leaflet, issues of the possible commercial proposals, proposed model forsharing the generated value).


16/100

That of external technical consultants: personnel from web agencies or technical companies,specialists of the field concerned by the innovation, they may have been met along the project inprofessional exhibitions; they are to authenticate the technical content of the innovation, discuss theconditions of its development (including giving pieces of advice on the necessary characteristics ofthe needed personnel), and be its first supporters by its future clients and commercial partners.Depending on the system achievement, this advice could be that of one big consultancy, specialisedin technology management and evaluation (eg, Gartner Group, Forrester Research, IDC, HurwitzGroup, Standish Group).

On top of answering each main question as to the real innovative content of the product, the evaluationteam will give advice on the content of each of the 20 documents or media necessary to the productbringing into the market:

1. Partners and partnership presentation2. Technology presentation3. Datasheets4. Technical white paper5. Return on investment6. Testimonials7. Competition analysis8. Evaluation package9. Demo application10. Rolling demonstration11. Installation guide12. Trouble shooting13. Getting started or methodology14. User’s manual15. Courseware16. Qualification questions17. Price list + configuration questions18. Frequently asked questions19. Proof of concept20. Executive summary

2.2.2 Measuring technology using financial measurements

Consideringtechno-economics, three criteria are mainly used as a toolbox to measure the value ofspecific technologies:

ÿ� Net Present Value (NPV) : the value today of cash received at a future date given aninterest rate.

Figure 2.2 - Net Present Value (NPV)

Year n

Value

100ÿ

117ÿ

Year 1 Year 2 Year 3

Interest Rate: 17%

136.89ÿ160.16ÿ


17/100

The goal of the investment or capital budgeting decision is to make investments that areworth more thanthey cost. The current “worth” of an investment is the present value (PV) of its expected future cashflows, and its net present value (NPV) is the difference between the investment’s present value and itscost, i.e., the amount by which an investment is worth more than it costs. The difficult part of computinga NPV is figuring out what an investment is worth, i.e., its present value.

ÿ� Payback Period (PP) : the time period needed before net savings equal initial cost.In most techno-economic studies, this criterion is considered as an excellent measure for riskestimation. It is also considered as a key measurement by portal consultant. Payback indicateswhen ROI = 0. Short payback periods focuses on an aggressive deployment strategy by thecompany.

Figure 2.3 - Payback period.

ÿ� Return on Investment (ROI): the average total savings over N years divided by the cost.

ROI = (Total savings year 1 + .... + Total savings year N) / Total cost over N years

Considering these three main financial measures, some basic rules must be respected to evaluate costs :

ÿ� Count everything that is directly associated with the project (I purchased a web server for thisproject),

ÿ� Don’t count infrastructure items not associated with the project (I used the existing web server),ÿ� Do count infrastructure items that were driven by the project. (The company purchased a web

server because of this project and two others like it - include 1/3 of the cost for a 3 years periodfor example).

These rules must be applied to some well-identified category of cost, such as:

ÿ� Software, ÿ� Consulting,ÿ� Hardware, ÿ� Training,ÿ� Personnel, ÿ� Other.

2.2.3 Human Assessment

The Goal-Question-Metric (GQM) technique [Nick, 1999] proposes the analysis of system performancefrom the user perspective. The GQM evaluation methodology includes the process, templates andguidelines for the application of GQM as proposed by [Briand et al., 1996]. The GQM methodology is anindustrial-proven technique that has been successfully used in the Software industry.

Time

Savings

Costs

Payback period


18/100

Figure 2.4- The basic principle of Goal Question Metric methodology

2.3 References

[Benjamin et al. 1999],Asuncion Gomez Perez, V. Richard Benjamins, "Overview of Knowledge, Sharing and Reuse

Components: Ontologies and Problem-Solving Methods". In Proceedings of the IJCAI-99 workshop on Ontologies and Problem-

Solving Methods (KRR5) Stockholm, Sweden, August 2, 1999. pp. 1.1--1.15

[Briand 1996], Briand L.C., Differding C.M., Rombach H.D. ,“Practical guidelines for measurement-based process

improvement”, Software process, 2(4): pages 253-280

[Cohen, 1995]Empirical Methods for Artificial Intelligence , MIT Press

[Grüninger et al.] Michael Grüninger, Mark S. Fox,"Methodology for the Design and Evaluation of Ontologies",Proceedings of

IJCAI'95, Workshop on Basic Ontological Issues in Knowledge Sharing, April 13, 1995

[Jasper et al. 1999] R. Jasper and M. Uschold, "A framework for Understanding and Classifying Ontology Applications ",

Proceedings of KAW'99

[Julieanne et al. 2000 a]Julieanne van Zyl, Dan Corbett,"Population of a Framework for Understanding and Classifying

Ontology Applications" ECAI'2OOO

[Julieanne et al. 2000 b]Julieanne van Zyl, Dan Corbett,"A Framework for Comparing the use of a Linguistic Ontology in an

Application", ECAI'2000

[Menzies, 1997] T. Menzies, "Evaluation Issues for Problem Solving Methods" , Banff KA workshop 1998

[Menzies, 1998] T. Menzies, "Evaluation Issues with Critical Success Metrics", Proceedings of KAW’98

[Menzies, 1998] T. Menzies, "Requirements for good measurements", http://www.cse.unsw.edu.au/~timm/pub/eval/cautions.html

[Nick 1999], Nick M., Althoff K., Tauz C. , “Facilitating the Practical Evaluation of Knowledge-Based systems and

Organisational Memories using Goal-Questions-Metric Technique”,.Proceedings of KAW99

[Rausher 1999] J. Rauscher, S. Marc, "A la conquête de la Silicon Valley" , Editions d’Organisation, Paris, ISBN: 2-7081-2282-7.

[Rombach, 1991], Rombach H.D. ,“Practical benefits of goal-oriented measurement”, in Fenton and Littlewood editors,

Software Reliability and Metrics, pages 217-235. Elsevier Applied Science, London.

[Sure et al. 1999] York Sure, Rudi Studer, "On-To-Knowledge Methodology - Employed and Evaluated Version", On-To-

Knowledge Project IST-1999-10132

� �ûú� ��ý��ú��ü�

� ��ü� ��þý�� þ�ý�

��

� � ��

�� üú� �� üú� � �� üú�

� � ��

�� þý��

�� ü�

��

��ý�ýþý��

��þ��þ�þý��


19/100

3 Most Representative Applications

3.1 IntroductionIn the field of Ontology-basedSystems, we have identified so far the following key clusters ofapplications:

ÿ� Corporate Intranet and Knowledge Management,ÿ� E-Commerce (B2B, B2C),ÿ� Information Retrieval,ÿ� Portals and Web communities.

e-commerce

KnowledgeManagement

Portals

InformationRetrieval

USER

E-COMMERCE

ÿ� Searching :o Products,o Serviceso Etc.

ÿ� Médiation Service,ÿ� Trading,

KNOWLEDGEMANAGEMENT

ÿ� Searchinginformation,

ÿ� Extractinginformation,

ÿ� Sharing information,M i i i

PORTALS

ÿ� Managing Services,ÿ� Knowledge sharing,ÿ� Integration/

development &Maintenance,

ÿ� Searching & accessing

INFORMATIONRETRIEVAL

ÿ� Searching,ÿ� Ranking,ÿ� Etc.

Text Images Other documents

PRODUCTS SERVICES

INDEX

ONTOLOGY

QUERYINTERFACE

KNOWLEDGEMANAGER

ONTOLOGYDESIGNER

WEB SERVICESLIBRARY

USER PERCEPTION

Improve customer service,Increase productivity,Improve collaboration,Cost effective & powerful systems,Lower network & storage costs, etc.

Recall and Precision,Coverage ratio,Novelty & Recall effort,Guidance in query formulation,Ranking of the retrieved documents,

Explaining the results,Robustness,Total ROI, etc.

Figure 3.1 Clusters of ontology-based applications with some additional evaluation criteria.


20/100

In each of those clusters, we provide:ÿ� An overall functional service definition of the application cluster,ÿ� An illustration through Main Use-Case2,ÿ� The selected most representative applications (2) or "killer application",ÿ� References for further information.

However, this classification could appear somewhat arbitrary or simply the mirror of the current mainintegrated applications making largely use of the Web technology. In fact when evaluating the IR mainclass of technology, we reached the conclusion that a global architecture might be proposed – see figurebelow – clearly distinguishing the Ontology part from the Reasoning (or inference) services as elementaryservices available to build all the known applications and future ones …

It is clear that we are on the current trends towards the Web services making full use this time of theSemantic Web technology.

In that version (D22) of the D2x series that aspect is just introduced, but if needed this could bedeveloped further in the following deliverables.

2 Use-Case diagramsare notations defined in UML, the “Unified Modeling Language” (largely used in Industrial Software

Engineering, and considered for extensions for the Knowledge Engineering emerging industry), which is basically a set of useful

representations. Typically they are used during the early phases of system engineering and show an abstraction level of the services

offered by the system. Use-Case put in the sceneactors who stand outside the system and who interact with the system through

services offered by the system.


21/100

3.2 Enterprise Portals and Knowledge Management

Knowledge Management is one important application where ontologies can play an important role.Ontologies can constitute a component of a corporate memory: they can then be explored by the end-userto discover the organization of the enterprise (e.g. enterprise ontology), or to study thoroughly a specifictechnical domain (e.g. domain ontology), etc.

Ontologies can also help to improve information retrieval through annotations of the resourcesconstituting the corporate memory: such semantic annotations can then allow ontology-guidedinformation retrieval.Several examples of KM recent applications strongly rely on ontologies and knowledge-based services.Let us cite:

ÿ� CoMMA project [Gandon, 2001; Gandon et al, 2002]: CoMMA offers an integrated solution toimplement a corporate memory based on ontologies and agent technology. It promotes a wide visionof the document retrieval issue that could be applied to several cases. To illustrate this approach, twoscenarios have been implemented through the project:

o Insertion of a new employee,o Technology monitoring.

CoMMA solution can be applied to several different contexts, related to the KM issue:

1. The memory is composed of heterogeneous evolving documents, structured using semanticannotations expressed with concepts and relations provided by a shared ontology.

2. The description of the different user groups, profiles and roles uses the concepts and relations ofthe ontology to make explicit, share and exploit a model of the organizational environment and userpopulation.

3. Ontology is not only a tool for document annotation and communication support, it is a fullcomponent of the memory highly relevant in itself.

ÿ� DECOR project [Abecker et al, 2002]: Having a document archive organized aroundontologicalstructures, the ontologies can be used to design knowledge portals for manual browsing,or they can be used by information retrieval algorithms in order to improve precision or recall whenevaluating queries. In DECOR project, business process models are considered as an ontologywhich can be used to specify the creation, or the potential usage context, for a given knowledge item

ÿ� FRODO project [Van Elst et al, 2002]: Ontologies facilitate access to, and reuse of knowledge inOrganizational Memories (OMs). In distributed OMs – as the next evolution step for practical appli-cations of OMs – we can no longer rely on the assumption of globally shared conceptualizations. Inorder to retain the benefits of domain ontologies the FRODO project proposes to explicitlycontrolthe sharing scopeof ontological knowledge. To this end are introduced ontology societies, whichare primarily defined by the rights and obligations of their members. The FRODO architecture forDistributed Organizational Memories is based on such ontology societies.

ÿ� SAMOVAR (Systems Analysis of Modeling and Validation of Renault Automobiles)[Golebiowska et al, 2001]aims at preserving the memory of the problems encountered during aproject in automobile design, so as, to reuse them in new projects. SAMOVAR relies on i) the semi-automatic building of ontologies by using a linguistic tool on a textual corpus, ii) the «semantic»annotations of the problem descriptions relatively to these ontologies, iii) the formalization of the


22/100

ontologies and annotations in RDF(s), and iv) the integration of the semantic search engineCORESE, that enables an ontology-guided search in the base of problem descriptions.

ÿ� KM-LIA project [López-García, 2001] : The Ontology Group of the Artificial IntelligenceLaboratory (LIA) - Technical University of Madrid, has built a KM application based on ontologies[López-García, 2001]. This application includes lessons learnt, news and search machines in thecontext of the LIA. KM-LIA allows multiple concurrent users, and it provides features ofinformation protection and management of permissions and users.

ÿ� This application has been completely integrated with the technology and systems belonging to LIA'sOntology Group, without duplicate resources. That is, KM-LIA has been built on the applicationserver called Minerva and the ontology access service of WebODE.

One of the most attractive features of KM-LIA is that you can easily adapt the application foranother environment different to the LIA.The system allows, in run-time, changing the currentontologies by others that model other institution.

3.2.1 Service definition

Nowadays, knowledge is one of the most crucial success factors for enterprises. Therefore, KnowledgeManagement (KM) has been identified as a strategically important means for enterprises. Clearly, KM isan interdisciplinary task, including human resource management, enterprise organization and culture aswell as IT technology. However, there is a widespread consensus that IT technology plays an importantrole as an enabler for many aspects of a KM solution.In the past, IT technology for knowledge management has focused on the management of knowledgecontainers using text documents as the main repository and source of knowledge. In the future, SemanticWeb technology, especially ontologies and machine-processable relational metadata, pave the way to KMsolutions that are based on semantically related knowledge pieces of different granularity: ontologiesdefine a shared conceptualization of the application domain at hand and provide the basis for definingmetadata, that have a precisely defined semantics, and that are therefore machine-processable. Although,first KM approaches and solutions have shown the benefits of ontologies and related methods, there stillexists a large number of open research issues that have to be addressed in order to make Semantic Webtechnologies a complete success for KM solutions:

ÿ� Industrial KM applications have to avoid any kind of overhead as far as possible. Therefore, aseamless integrationof knowledge creation, e.g. content and metadata specification, andknowledge access, e.g. querying or browsing, into the working environment is required.Strategies and methods are needed for supporting the creation of knowledge, as side effects ofactivities that are carried out anyway. This requires means foremergent semantics, e.g. throughontology learning, which reduces the overhead of building-up and maintaining ontologies.

ÿ� Access to as well as presentation of knowledge has to becontext-dependent. Since the contextis set-up by the current business task, and thus, by the business process being handled, a tightintegration of business process management and knowledge management is required. KMapproaches being able to manage knowledge pieces provide a promising starting point for smartpush services that will proactively deliver relevant knowledge for carrying out the task at hand.

ÿ� Conceptualization has to be supplemented bypersonalization. Taking into account theexperience of the user and his/her personal needs is a prerequisite for avoiding informationoverload, on the one hand and for delivering knowledge on the right level of granularity, on theother hand.


23/100

The development of knowledge portals serving the needs of companies or communities is still a more orless manual process. Ontologies and related metadata provide a promising conceptual basis for generatingparts of such knowledge portals. Obviously, conceptual models of the domain, the users and the tasks areneeded among others.Generation of knowledge portals has to be supplemented with the (semi-)automatic evolution of portals. Since business environments and strategies change rather rapidly, KMportals have to be kept up-to-date in this fast changing environment. Evolution of portals also includes theaspect of ‘forgetting’ outdated knowledge.

KM solutions will be based on a combination of intranet-based functionalities and mobile functionalitiesin the very near future. Semantic Web technologies are a promising approach to meet the needs of themobile environments, like e.g. location-aware personalization and adaptation of the presentation to thespecific needs of mobile devices, i.e. the presentation of the required information at an appropriate levelof granularity. In essence, employees should have access to the KM applicationanywhereandanytime.

Peer-to-Peer computing (P2P),combined with Semantic Web technology, will be an interesting path toget rid of the more centralized KM solutions that are currently implied by ontology-based solutions. P2Pscenarios open up the way to derive consensual conceptualizations among employees within an enterprisein a bottom-up manner.

Virtual organizations become more and more important in business scenarios, mainly due todecentralization and globalization. Obviously, semantic interoperability between different knowledgesources, as well as trust, is necessary in inter-organizational KM applications.

The integration of KM applications (e.g. skill management) withe Learning is an important field thatenables a lot of synergy between these two areas. KM solutions and e Learning must be integrated fromboth an organizational and an IT point of view. Clearly, interoperability and/or integration of (metadata)standards are needed to realize such integration.

Knowledge Management is obviously a very promising area for exploiting Semantic Web technology.Document-based KM solutions have already reached their limits, whereas semantic technologies open theway to meet the KM requirements of the future.

3.2.2 Use Case and Needs

The scenario shown in figure 3.2, builds on the distinction betweenknowledge process(handlingknowledge items) andknowledge metaprocess(introducing and maintaining KM systems).Ontologies constitute the glue that binds knowledge sub-processes together. Ontologies open the way tomove from a document-oriented view of KM to a content-oriented view, where knowledge items areinterlinked, combined, and used.The scenario shows that you can clearly identify and handle different sub-processes that drive thedevelopment and use of KM applications. You support these sub-processes by appropriate tools that aretied together by the ontology infrastructure.


24/100

Figure 3.2- Example of Knowledge Management Use Case.

3.2.3 Business-to-Employee (B2E) as a most Representative Applications

One of the key applications or may-beKiller Application of corporate Intranet is its usage in theMobilee-businesshas the potential to deliver business-to-employee interactions capable of creating significantcompetitive edge. The real-life advantages of mobile access to business critical applications require thatall organizations, including technology builders, have strategies in place for interacting with their ownemployees.

[Dyson, IBM 2001] Thebusiness-to-employee (B2E)opportunity will be the starting place for the real,extended application of mobile e-business. Mobile e-business is a natural progression of e-business –delivering applications and services to employees on the move. Importantly, mobile e-business is not onlyfor companies selling consumer leisure gadgets, but also for organizations seeking to achieve addedproductivity, increased profitability, improved competitiveness and optimal return on investment.The key to making the most of mobile e-business for B2E interactions is connecting the infrastructure andmobile device to line-of-business systems – Intranet collaboration, client billing information, clientdatabases, production lines, e-mail, calendars and customer relationship management applications. Inshort, providing personalized and specific ways to take advantage of company information and resources.

For more information dedicated to some representative applications for KM, reader is invited to consultannex 1.

KnowldegeManager

Knowledge Worker

Nav. / Browse KBSeek Knowledge

Querying KB

Com. Know. Sharing

Push Services

Ontology Building

KnowledgeEngineer

Maintenance

Annotation

KnowledgeProvider

Fill KB


25/100

3.2.4 References[Abecker et al., 2002] A. Abecker, A. Bernardi, S. Dioudis, L.Van Elst, R. Herterich, C. Houy, M. Legal, G. Mentzas, S. Müller, G.

Papavassiliou.Enabling Workflow-Embedded On Access With The Decor Toolkit.To appear in R. Dieng-Kuntz and N. Matta

eds, Knowledge Management and Organizational Memories, Kluwer, 2002.

[Van Elst, 2002] L. Van Elst, A. Abecker,Domain Ontology Agents In Distributed Organizational Memories. To appear in R.

Dieng-Kuntz and N. Matta eds, Knowledge Management and Organisational Memories, Kluwer, 2002.

[Gandon, 2001] F. Gandon,Engineering an Ontology for a Multi-Agents Corporate Memory System, Proc. of the Eigth

International Symposium on the Management of Industrial and Corporate Knowledge, Technological University of Compiègne,

France, 22-24 October 2001.

[Golebiowska et al, 2001] J. Golebiowska, R. Dieng-Kuntz, O. Corby, D. Mousseau,Samovar : Using Ontologies And Text-

Mining For Building An Automobile Project Memory , Proc. of K-CAP, Victoria, October 2001.

[Staab et al., 2001], Steffen Staab, Rudi Studer, Hans-Peter Schnurr, York Sure,“Knowledge Management : Knowledge

Processes and Ontologies”; IEEE Intelligent Systems Journal, Jan. Feb. 2001.

[Dieng et al., 1999], R. Dieng, O. Corby, A. Giboin, and M. Ribiere.Methods and tools for corporate knowledge management.

Int.Journal of Human-Computer Studies, 51(3):567–598, 1999.

[Kuehn et al., 1997], O. Kuehn and A. Abecker.Corporate memories for knowledge memories in industrial practice: Prospects

and challenges. Journal of Universal Computer Science, 3(8), August 1997.

[Schreiber et al., 1999], G. Schreiber, H. Akkermans, A. Anjewierden, R. de Hoog, N. Shadbolt, W. Van de Velde, and B. Wielinga.

Knowledge Engineering and Management — The CommonKADS Methodology. The MIT Press, Cambridge, Massachusetts;

London, England, 1999.

[Dyson S. IBM, 2001], S. Dyson, IBM UK, «B2E Mobile e-Business : Driver, Passenger or Spectator ?», in [E-work and E-

commerce 2001].

[D. fensel, 2001] D. Fensel: "Ontologies: Silver Bullet for Knowledge Management and Electronic Commerce". Springer-

Verlag, Berlin, 2001.

[D. Fensel et al. 2002] D. Fensel, J. Hendler, H. Lieberman, and W. Wahlster (eds.): "Semantic Web Technology", MIT Press,

Boston, to appear.

[D. Fensel, 1998], D. Fensel, S. Decker, M. Erdmann, and R. Studer: "Ontobroker in a Nutshell" (short paper). In C. Nikolaou et

al. (eds.), Research and Advanced Technology for Digital Libraries,Lecture Notes in Computer Science, LNCS 1513, Springer-

Verlag Berlin, 1998.

[O'Leary, et al. 2001] Dan E. O’Leary and Rudi Studer (eds.): "Knowledge Management", IEEE Intelligent Systems 16 (1),

January/February 2001 (special issue).

[Staab S. et al. 2001] Steffen Staab and Alexander Maedche, "Knowledge Portals — Ontologies at Work". AI Magazine21(2),

Summer 2001.

[López-García, 2001] Asun López-García , "KM-LIA ", Final year project report of the Artificial Intelligence Laboratory (LIA),

Technical University of Madrid, 2001.

IST Knowledge Management portal, http://www.knowledgeboard.com/


26/100

3.3 E-Commerce

3.3.1 Why are ontologies promising for e-Commerce?

At the present time, ontology and more generally ontology-based systems, appear as acentral issueforthe development ofefficient andprofitable Internet commerce solutions. They represent a way to accesswith efficiency and optimization to a large scale of Internet information (professional, business, leisure,etc.) spaces, which will be more and more prominent and determining feature of most business,governmental and personal informational activity in the near future. However, because of an actual lackof standardization for business models, processes, and knowledge architectures, it is today difficult forcompanies to achieve the promisedreturn on investment (ROI) from the e-commerce.

Moreover, it also exists a technical barrier that delay the emergence of e-commerce, lying in the need forapplications tomeaningfully share information, taking into account the lack of reliability and securityof the Internet. This fact may be explained by the variety of enterprise and e-commerce systems deployedby businesses and the way these systems are variously configured and used. As an important remark, suchinteroperability problems become particularly acute when a large number of trading partners attempt toagree and define the standards for interoperation, which is precisely a main condition formaximizing theReturn on Investment (ROI).

Although it is useful to strive for the adoption of a single common domain-specific standard for contentand transactions, such a task is still often difficult to achieve, particularly in cross-industry initiatives,where companies co-operate and compete with one another.

In addition to this:

ÿ� Commercial practicesmayvary in a wide way and consequently, cannot always be aligned fora variety of technical, practical, organizational and political reasons.

ÿ� The complexity of the global descriptionof the organizations themselves: their products and/orservices (independently or in combination), and theinteractions between them remain aformidable task.

ÿ� It is usually very difficult to establish, a priori, rules (technical or procedural) governingparticipation in an electronic marketplace.

ÿ� Adoption of asingle common standard may limit business models, which could be adopted bytrading partners, and then, potentially reduce their ability to fully participate in Internetcommerce.

Because of all these aforementioned reasons, ontologies appear as really promising for e-commerce.Indeed, alternative strategies may consist of sharing foundational ontologies, which could be used as thebasis for interoperation among trading partners in electronic markets. An ontology based approach has thepotential to significantly accelerate the penetration of electronic commerce within vertical industrysectors, byenabling interoperability at the business level, reducing the need for standardisation at thetechnical level. This will enable services to adapt to the rapidly changing online environment.


Electronic Commerce is mainly based on the exchange of information between involved stakeholdersusing a telecommunication infrastructure. There are two main scenarios: Business-to-Customer (B2C)and Business-to-Business (B2B).


27/100

B2C applicationsenable service providers to propagate their offers, and customers to find offers, whichmatch their demands. By providing a single access to a large collection of frequentlyupdated offers andcustomers, an electronic marketplace can match the demand and supply processes within a commercialmediation environment.

B2B applications have a long history of using electronic messaging to exchange information relating toservices previously agreed between two or more businesses. Early plain-text telex communicationsystems were followed by electronic data interchange (EDI) systems based on terse, highly codified, wellstructured, messages. Recent developments have been based on the use of less highly codified messagesthat have been structured using the eXtensible Markup Language (XML).A new generation of B2B systems is being developed under the ebXML (electronic business in XML)label. These will use classification schemes to identify the context in which messages have been, orshould be, exchanged. They will also introduce new techniques for the formal recording of businessprocesses, and for the linking of business processes through the exchange of well-structured businessmessages. ebXML will also develop techniques that will allow businesses to identify new suppliersthrough the use of registries that allow users to identify which services a supplier can offer.

The coding systems used in EDI systems are often examples of limited scope, language independent,mini-ontologies that were developed in the days when decimalised hierarchical classification systemswere the most sophisticated form of ontology. There is a strong case for the redesign of many of theseclassification schemes based on current best practice for ontology development. ebXML needs to includewell managed multilingual ontologies that can be used to help users to match needs expressed in theirown language with those expressed in the service providers language(s).

Within Europe many of the needs of B2C applications match those of B2B applications. Customers needto use their own language to specify their requirements. These need to be matched with services providedby businesses, which may be defined in languages other than those of the customer. Businesses may ormay not provide multilingual catalogues. Even where multilingual catalogues are supplied they may notcover all European languages. For a single market to truly exist within Europe it must be possible forcustomers to be able to request product and sales term information in their own language, possiblythrough the use of on-line translation services. It is anticipated that services providing multilingualsearching of sets of catalogues will act as an intermediary between businesses and their potentialcustomers (figure 3.3)

3.3.3 Use Cases and Needs in B2CFor B2C transactions customers need access to electronic systems that:

ÿ� Provide a user interface in the customer's language.

ÿ� Provide access to details of products from a range of different firms that create products of the typerequired. For adequate comparison of products their descriptions will need to be “standardized” sothat they include comparable qualities.

ÿ� Allow users to request searches of multiple supplier catalogues using natural language terms so thatthe system can help them to identify which suppliers supply relevant products. For adequateresponse to search requests search engines will need to either use “standardized” sets of productdescription properties or use ontologies that provide facilities for identifying equivalences betweenterms.

ÿ� Provide users with a means of paying for the product that is acceptable and accessible to them andwhich the supplier can also accept. (This may require the services of one or more third party servicesuppliers, either in the form of credit card or banking services, or of “mall operators” willing to actas debt intermediaries.)


28/100

ÿ� Provide means for the timely and properly managed delivery of products from supplier to customer,including facilities for the automatic identification of most appropriate supply point for therequested delivery point.

Figure 3.3 – Example Use case and needs in B2C and B2B

3.3.4 Use Cases and Needs for B2B

The following needs exist:

ÿ� Trading partners will need to be able agree on the set of terminology that will be used to describe theproducts to be traded with respect to their own domains. (For example, a Dutch adhesivemanufacturer supplying wood glue to an Italian furniture maker will need to use terms and propertysets that are relevant to Italian furniture makers to describe their products.)

ÿ� Need to assign the labels used to identify specific fields within a message in a form that is dependenton locale and domain specific terminology (e.g. the terms used to describe transportation withinFinland, or the terms used to describe viniculture in Hungary)

ÿ� Need to be able to express the options provided in choice lists as locale-specific terms those creatingor receiving messages can understand in their own language

ÿ� Need to use terms and conditions that have previously been translated into the language of thebusiness accepting the terms (which could be the supplier or the customer in the B2B transaction)

ÿ� Need to adjust the legal and financial constraints on transactions to take account of the source ofsupply and the point of delivery. (For example, a company in Greece could ask a company inPortugal to ship goods from its Swedish warehouse to a delivery point in Denmark. Whilst thecontract may be between Greek and Portuguese company, in terms of the laws of one of thesecountries, the financial information used for customs purposes must be expressed in termsacceptable to the Danish and Swedish authorities, in the currencies they are currently using.)

M a n a g er

(f ro m A c to rs)

M a n a g e m e n t

(f ro m M a n a g e m e n t )

M u lt i l in g u a l Tra d in g

(f ro m M u l t i l i n g u a l T ra d i n g )

U s e r s u b s c rip t io n t o t h e S y s t e m

(f ro m A c c e ss)

C ro s s - l in g u a l In fo rm a t io n R e q u e s t

(f ro m C ro ss L i n g u a l )

C P K n o w le d g e m a n a g er

(f ro m A c to rs)

E n d - u s er

(f r om A c to rs )

M u lt i l in g u a l c a t a lo g u in g

(f ro m M ul ti li n g u al C a ta l o g u in g )

A c c e s s t o t h e S y s t e m

(f ro m A c c e ss)

C a t a lo g u e r

(f ro m A c to rs)


29/100

3.3.5 Representative Applications using Ontologies for E-Commerce

The following uses for ontologies, and classification schemes that could be defined using ontologies, havebeen noted within electronic commerce applications:

ÿ� Categorization of products within cataloguesÿ� Categorization of services (including web services)ÿ� Production of yellow page classifications of companies providing servicesÿ� Identification of countries, regions and currenciesÿ� Identification of organizations, persons and legal entitiesÿ� Identification of unique products and saleable packages of productsÿ� Identification of transport containers, their type, location, routes and contentsÿ� Classification of industrial output statistics.

Many existing B2B applications rely on the use of coded references to classification schemes to reducethe amount of data that needs to be transmitted between business partners. Such references overcome theproblems introduced by the natural ambiguity of words that have more than one meaning (polysems) orcan apply to more than one object (e.g. personal names such as John Smith). By providing a separate codefor each different use of the term it is possible to disambiguate messages to a level where they can behandled without human intervention.

Very few of the existing classification schemes used within electronic commerce applications have beendefined as formal ontologies, or have been formally modeled to ensure that the relationships betweenterms are fully described. To date most of the techniques introduced by ontologies have been applied togeneral linguistic situations, such as those involved in specific academic disciplines, rather than to thelanguage adopted by specific industries.

3.3.5.1 IST-MKBEEM

The MKBEEM platform focuses on adding multilinguality to the following stages of the informationcycle for multilingual B2C portal services: products or services content and catalogue semi-automatedmaintenance; automated translation and interpretation of natural language user requests, and naturaldialogue interactivity and usability of the service making use of combined navigation and naturallanguage inputs.The main overall goals of MKBEEM are to: develop intelligent knowledge-based multilingual keycomponents (NLP and KRR) for applications in a multilingual electronic commerce platform; validateand assess the prototypes on a pan-European scale (France and Finland) with three basic languages(Finnish, English and French) and two optional languages.

3.3.5.2 CEN/ISSS MULECO

The Multilingual Upper-Level Electronic Commerce Ontology (MULECO) currently being proposed bythe CEN/ISSS Electronic Commerce Workshop is designed to provide a mechanism wherebyrelationships between the high-level terms in business ontologies can be inter-connected. Recognizingthat most existing ontologies are domain and language specific, and that there is a need to be able to relateterms in one ontology with those in another ontology as part of the semantic translation that is needed tointer-relate applications (as defined within the E-Commerce Integration Meta-Framework (ECIMF) beingdefined by the CEN/ISSS EC Workshop), the MULECO team are hoping to define techniques that willallow multilingual querying of ontologies based on the relationships between the local ontologies and aset of well defined business and business process classification schemes.

3.3.5.3 IST-SmartEC

The goal of Smart-EC is to build the e-Trade platform on which the buyer will be able to purchase aGlobal Service (of the user’s demand requires several providers and involves more than one actors) as asingle service. During the first stage of the negotiation Smart-EC will help the user defining his/herdemand and will propose one or more global and convenient solutions. After the user agreement, Smart-


30/100

EC will be in charge of confirming the transaction to the respective providers and to distribute thepayment to the respective third parties.

The Global Service will be broken in more atomic services according to the knowledge modelled byOntologies. The Ontology rules will modelled the generic / specific relations between services, the timeand coherence constraints between services. Rules will also specify the commercial behaviour of servicessuch as the necessity to check availability on-line.

3.3.5.4CHEMDEX [content aggregation]

Chemdex Corporation is one of the main providers of electronic commerce solutions to the life scienceresearch industry. Chemdex is as a "vortex" business, essentially a Web-based market maker that bringstogether a fragmented group of buyers with an equally fragmented supply chain with a richness ofproduct information and inconsistencies in the technical specification.

The worldwide market for life science research supplies is more than $10 billion annually, including morethan $4 billion spent on biological and chemical reagents. Traditionally, products used to be offeredthrough paper catalogs from which scientists select the products desired and place orders with multiplesuppliers. Many research products are highly technical, requiring a large amount of supplemental productinformation to assist in the purchase decision. This structure makes the purchasing process inefficient andhas created the need for a Chemdex type approach.

Chemdex.com is a fully transactional, Web-based electronic catalog and ordering system for life scienceproducts. Using chemdex.com, buyers are able to search for products from multiple suppliers, review allthe technical product information available for each product, and then make an informed purchasingdecision.

Figure 3.4 -The Chemdex marketplace

At Chemdex.com, the content management team converts the entries from multiple catalogues into acommon ontology. Domain experts are crucial to this process. The ontology development process beginswith an analysis of supplier-side ontologies and buyer-side behavior. Once candidate ontology emergesfrom this analysis, it is used to drive both the content engineering process and the design of appropriatesearch tools. Content and search are reunited when users interact with the system. With time and with


31/100

usage, the ontology evolves and stabilizes, eventually reaching wide industry adoption when the valueand opportunity present within the marketplace rises rapidly.

The Chemdex marketplace relies upon a shared ontology of life science products. The ontology permitsdifferent product classification schemes to be accommodated within a representation which is acceptableto each of the participating trading partners.

The main role and benefits of the use of ontologies within the Chemdex solution is that ontologies :ÿ� Allow a common information structure for all the players in this market,ÿ� Promote a shared language that de-fragments both the demand and the supply side,ÿ� Facilitate product comparison,ÿ� Support a simpler user interface.

3.3.5.5 ALICE

ALICE system [Carvalho et al., 2001] is under development both by the Knowledge Media Institute(KMi) of the Open University of United Kingdom and the Icelandic Internet Company, INNN. The aimof Alice is to construct a personal Web shopping assistant that uses a variety of knowledge sources tocreate a personalized shopping experience. Alice can guess characteristics of the clients, and it helps themto choose the best products according to such characteristics. It also shows to each client the goods onoffer that are most appropriate for him (her). The idea is to recover the "local corner shop" for Internet.

3.3.6 References

[Carvalho et al., 2001] De-Carvalho M., Domingue J., Pertusson H., "Alice: An ontology based architecture for supporting

online shopping", K-CAP2001- Workshop Knowledge in E-Business. Victoria, Canada, 2001

Web :http://kmi.open.ac.uk/projects/alice/

[MKBEEM 2000]

URL: http://www.mkbeem.com/

[MULECO 2001]

URL : http://www.cenorm.be/isss/Workshop/ec/MULECO/Documents/Muleco_Documents.htm

[SmartEC, 2000]

URL : http://www.telecom.ntua.gr/smartec/

[E-work and E-commerce 2001] "Novel solutions and practices for a global networked economy", e.2001, 17-20 October 2001,

Venice, Italy. Edited by Brian Stanford-Smith and Enrica Chiozza, ISBN: 1 58603 205 4.

URL: http://www.ebew.net/programme.htm

[eWork 2001], "Status Report on New Ways to Work in the Knowledge Economy", IST document

URL: http://www.eto.org.uk


32/100

3.4 Information Retrieval

The last few years have seen a dramatic increase in the generation of information as a result of highthroughput technologies in certain leading edge fields (e.g. in the genomics field with the reading of thehuman genome) but also as a consequence of the ease with which information can be published. At thesame time B2B integration creates a need for the interoperability of IT resources and consequently forsystems that will support this in an effective manner.

A result of this explosion is that theextraction of knowledgefrom information, andnot the availabilityof information itself, has become acritical factor for business. From a business perspective therefore,effective information retrieval has become an essential element of the competitive capability ofcompanies within knowledge-intensive industries. Furthermore, the central role that information plays indecision-making creates the need for high quality of the retrieved information.

Most technologies that have been used to date have paid lip service toquality issuespartly because otherperformance metrics (such as response times) were of higher priority. This however is no longer the caseand instead quality issues (such asrelevance, accuracy, completeness, conciseness) are now important.

Ontology-based approaches promise to increase the quality of responses since they aim to capture withincomputer systems some part of the semantics of concepts allowing for better information retrieval.However, while the opportunities for value creation exist, the path is anything but straight.


In an information retrieval (IR) application, ontologies are used to guide the search so that the systemmay return more relevant results. The assumption in this class of application is that the ontology willallow the IR system a better representation (“understanding”) of the concepts being searched and thusmake possible an improvement of its performance from what is presently the case.

The problems of IR are well known to the research and user communities. Amongst the most widelyrecognized ones are the so-calledmissed positivesandfalse positives. In the first case the system fails toretrieve relevant answers to the query whereas in the second case the system retrieves answers that areirrelevant to the query. Throughout the years a number of mechanisms of various levels of sophisticationhave been devised for ranking the results produced from the search. These mechanisms range fromhuman based indexing, to statistical measures (word frequency analysis) and more recently to Internet-related measures such as page popularity and number of incoming links.

While the human-based method might be considered the best for most cases, the rapid increase in theamounts of information presently being generated makes this an increasingly untenable proposition. Theuse of ontology-enhanced search and retrieval promises to address this issue since it attempts to replicatethe level of quality of human-based representation of information and concepts while still being able tohandle vast amounts of data.However, even such an approach risks failing if it does take into account generic search issues that havenot been tackled in the past, as well as some issues that are specific to search through the Internet. Themain issues are:

ÿ� Context: context is the conceptual framework that determines how relevant a query result is. Contextis an issue that has not been successfully addressed to date. However, context can be said to bedetermined by one or more of the following:

o Individual person profile,o ‘Mode’ of the search (whether it is search, discovery or browsing),o Domain of interest,


33/100

o Goal of the search,o Business process,o Organizational model,o Etc.

ÿ� Information quality : This refers to issues such as how up to date the information is, the existence ofdifferent versions of the same information as well as the existence of contradictory information

ÿ� User search mode: the goal of a search may often differ ranging from browsing, to discovery, tofocused search. In each case the user will be expecting increasingly relevant results, which has adirect bearing on the search and ranking mechanisms employed.

3.4.2 Use Cases and Needs

The following figure 3.5 illustrates an application scenario that addresses the issues mentioned above.Ontology creation and maintenance are not indicated. On the other hand the architecture showndistinguishes between internal (proprietary) and external (non-proprietary) sources indicating the need forontology alignment in the latter case.

Note alsothe central role of the ontologyboth in the ‘query augmentation’ phase and in the processingof the initial results. In the former case the ontology will form the basis for better understanding thecontext of the user’s initial query taking into account the search mode employed.

The ontology is also used in thefiltering, ranking and presentation of the results covering qualityissues such as contradictions and related information (a different possible answer to the same query or ananswer to a different but related query).

Figure 3.5 - Basic architecture for ontology based information retrieval applications

INTERNAL

ONTOLOGY

QUERY

SEARCHMODE

CONTEXT

RANKINGMODULE

INTERNAL

DATABASE

SEARCHENGINE

QUALITYANALYSIS

Timeliness,Contradictions,

Versions

AUGMENTEDQUERY

INITIALRESULT

QUERYRESULT

EXTERNAL

ONTOLOGY

EXTERNAL

DATABASE

ONTOLOGYALIGNMENT


34/100

3.4.3 Representative Applications

To date a large variety of ontology-based applications have been developed that focus on the access,organization, exchange and aggregation of information. These applications can be grouped in two mainclasses: tools and domain-specific applications.

By ‘tools’ we mean software systems that aim to create, maintain and in general manage ontologies; theseare discussed in Section 6. In the following paragraphs we present 5 representative applications that coverthe following tasks that are related to Information Retrieval:

1. Content Management: refers to the categorization/indexing of documents and other source data.

2. Query Augmentation: refers to the expansion of the user query so as to better understand thecontext in order to return more relevant results

3. Content Aggregation/presentation: refers to the presentation of content to the user and coversboth the collection and integration of content from various sources (increasingly made possibleby the www and the supporting technologies) and the creation of intuitive user interfaces.

3.4.3.1 AUTO-CATEGORIZER [ content management]

Auto-Categorizer is an information management and retrieval tool developed by Applied Semantics, Inc.,which features a taxonomy administration tool to enhance its content categorization application. The toolprovides users the ability to quickly import, create and edit sets of categories within a variety of industry-standard and user-defined taxonomies, allowing enterprises to build and maintain an informationhierarchy that mirrors the knowledge hidden in their data repositories.Typically, information managers or taxonomists must gather dozens, sometimes hundreds, of documentsrelated to a specific category in order to attach meanings and associations to that category when usingother automatic categorization technologies. The ontology-based technology used in Auto-Categorizermaps concepts in taxonomies related to specific business objectives supporting content categorization in adirect manner. Auto-Categorizer works in real-time and can organize documents into an industry standardor custom taxonomy.

Unlike technologies that require creating a training set of documents or designing a set of rules to makethe categorization system function, Auto-Categorizer allows the user to define category names andassociate categories with the concepts in the ontology that best describe the category.

3.4.3.2 ARCH [query augmentation]

ARCH (Adaptive Agent for Retrieval Based on Concept Hierarchies) [Bamshad 2001] is a client-sideagent, for assisting users in formulating an effective search query. The agent utilizes a hierarchicallyorganized semantic knowledge base in aggregate form, as well as an automatically learned user profile, toenhance user queries. In contrast to traditional methods based on relevance feedback, ARCH assists usersin query modification prior to the search task.

The initial user query is (semi-) automatically modified based on the user's interaction with an embedded,but modular, concept hierarchy. The modular design of the agent allows users to switch among therepresentations of different domain-specific hierarchies depending on the goals of the search.

ARCH passively learns a user profile by observing the user's past browsing behavior. The profiles areused to provide additional context to the user's information need represented by the initial query. The fullsystem also incorporates mechanisms for categorizing and filtering the search results, and using thesecategories for performing refined searches in the background.


35/100

Preliminary experiments have shown that the agent can substantially improve the effectiveness ofinformation retrieval both in the general context of the Web, as well as for search against domain-specificdocument indexes.

3.4.3.3 MEDICAL CONCEPT MAPPER [query augmentation]

The Medical Concept Mapper is a tool, which combines the Unified Medical Languages System (UMLS)developed by the National Library of Medicine (NLM) with automated computational search space toolsdeveloped by the AI Lab at the University of Arizona. The Medical Concept Mapper consists of fivecomponents that combine human-created and computer-generated elements. The Arizona Noun Phraserextracts phrases from natural language user queries. WordNet and the UMLS Metathesaurus providesynonyms. The Arizona Concept Space generates conceptually related terms. Semantic relationshipsbetween queries and concepts are established using the UMLS Semantic Net.

Medical Concept Mapper is used as an aid for providing synonyms and semantically related concepts toimprove searching. All terms are related to the user-query and fit into the query context. The system is anin-depth integration of manually created ontologies and computer generated tools, the intertwining ofwhich allows a synergy to surface that surpasses the weaknesses and strengths of each tool when used onits own.

MCM has been tested on two expert groups: medical librarians and cancer researchers.The test showed that the system can automatically double the number of useful search terms extractedfrom queries. It can also suggest related terms with high precision.

3.4.3.4SARI - A SYSTEM FOR SEMANTICAL INFORMATION RETRIEVAL [content aggregation]

The SARI (Software Agents for Retrieval of Information) system is intended to act as a broker betweenhuman users or other computerized systems (i.e. applications) needing information at one end, andheterogeneous information sources with different search engines at the other.SARI’s architecture reflects the system’s role as a broker between its users and information sources.SARI’s has agents of the following types:

ÿ� Application Agents represent the users (humans or other computerized systems) to the SARIsystem.

ÿ� Search Agents mediate information sources. They compile queries coming from Control Agentsinto the query languages of their information sources, and send the results back to the ControlAgents.

ÿ� Control Agents act as brokers in the SARI system. Each Control Agent receives agent messagescontaining information retrieval requests from Application Agents, decides to which SearchAgents it forwards the requests, sends messages containing the retrieval requests to theappropriate Search Agents, receives messages containing search results from the Search Agents,combines them into information retrieval results, and sends the retrieval results back to theApplication Agents.

ÿ� Ontology Agent contains metadata in the form of ontologies that describe the conceptualstructure of the information present in the information sources used by SARI.

In addition, there are also Content Provider Agents that represent content providers to the SARI system.Content providers are organizations or individuals who own one or more information sources that areaccessible to the SARI system. Control Agents form the heart of SARI. They make their brokeringdecisions on the grounds of the user information lying in user profiles, and of the metadata about theinformation to be retrieved lying in ontologies. Control Agents can form federations with each other, as arule, but there is just one Control Agent in the present pilot version of SARI.


36/100

The conceptual structure of the information contained in the information sources available to SARI isdescribed by ontologies. The ontologies describing Web resources are specified as RDF schemas anddescriptions for SARI. Ontologies can be graphically browsed in SARI.

Figure 3.6 - Architecture of SARI.

In SARI the concepts of different ontologies are linked to each other by making use of the notions ofviewpoint and bridge. The ontologies interlinked in such a way form the ontological structure that can beviewed from different perspectives. For example, there is a bridge between the concepts Commodity andproduct, which are respectively the root classes of the classifications under the foreign trade, andmanufacturing viewpoints.Future goals with SARI include making the formation of bridges between the concepts of differentontologies semiautomatic, and also semiautomatic generation of RDF metadata from Web resources.

VTT Information Technology, Tampere University of Technology, and Tampere University have beenworking out the SARI system in Finland, jointly. The project started in March 1996, and ended March1999.

3.4.3.5 OntoSeek

For Yellow pages and products catalogues, structured content representations coupled with linguisticontologies can increase both recall and precision of content-based-retrieval. The OntoSeek system adoptsa language of limited expressiveness for content representation and uses a large ontology based onWordnet for content matching.

OntoSeek combines an ontology-driven content-matching mechanism with moderately expressiverepresentation formalism. Differently from most of current systems, the user is not assumed to havefamiliarity with the vocabulary used for component encoding, but the system relies on a large linguisticontology called Sensus to perform the match between queries and data. It assumes that the informationencoding and retrieval processes will involve a degree of interactivity with a human user.


37/100

3.4.3.6 Semantic Miner: Smarter Knowledge RetrievalIntroduction. The SemanticMinerTM3 (cf. figure 3.7) is a Knowledge Retrieval platform that combinessemantic technologies with conventional retrieval approaches. Theimproved navigationenables the userto easily define semantic queries to all kinds of information sources – especially unstructured documents.Semantic information integrationallows fordifferent viewsand deep analysis of hidden knowledge by theexternalization of implicit knowledge.

Architecture. SemanticMinerTM is designed in client-server architecture. It provides informationretrieval in various data sources (e.g. files indexed with an index-server, hypertext-pages reached with aWWW search-engine, and data stored in a database via a DBMS). The SemanticMinerTM -Server (SMS),which is a specialized OntoBrokerTM -system [Fensel et al., 2000], provides the interface to the datasources as well as the inference engine to retrieve and present implicit knowledge. Zope serves as webserver to provide the client-side web-interface and as application server to query theSMS. This flexiblemiddleware architecture allows the SemanticMinerTM to become easily administrative and configurablethrough one central interface. OntoEditTM [Sure et al, 2002] provides collaborative design, adaptation andimport of ontologies as knowledge models to feed the SMS.

Improved Navigation. The use of ontologies provides a simplified navigation. On the one hand the usergets easy access to relevant information by browsing through the modeled concepts and their relations.On the other hand the use of synonym lists and thematically classification guides the user automatically torelevant search items.

Different Views. The use of multiple ontologies allows the SemanticMinerTM to provide different viewson the same content respectively information.

Figure 3.7 - SemanticMiner

Semantic Information Integration. Through the combination of a search request as textual informationwith structured information (e.g. lists, databases, meta data) and logical rule cohesion the performance ofthe SemanticMinerTM approach is further increased. The overall goal is to provide essential knowledgecontents instead of links to documents containing the content. An example is the combination of a searchrequest with a list of employees, which could be taken from an arbitrarily source (e.g. a human-resources-system). As a result to the request for an expert the user receives not only a list of documents but a rankedselection of experts for the specified topic. This is achieved through either the use of meta data or theformation of collocations. Another search request would be the combination of subject areas withprojects.

3SemanticMinerM3 is available from http://www.ontoprise.com


38/100

Externalize Knowledge.Additional benefit originates in the appraisal of the logical rule cohesions withthe inference engine to retrieve and present implicit knowledge.

Conclusion. As shown above, SemanticMinerTM provides quicker, better, and smarter knowledgeretrieval.

3.4.3.7 Other applications

Other interesting applications are the ones derived from theAriadne andOBSERVER projects.

The main purpose of the Ariadne project (http://www.isi.edu/info-agents/ariadne/index.html) is thedevelopment of technology and tools for rapidly constructing intelligent agents to extract, query, andintegrate data from web sources. It is being developed in the Information Sciences Institute (ISI) at theUniversity of Southern California, with the support of DARPA. This project tries to improve SIMS[Arens et al.; 1996] allowing the management of semi-structured sources such as Web pages.

The Ariadne approach is being used in a country information agent that combines data about the countriesin the world from a variety of information sources [Knoblock et al.; 2001]. It is also being used in asystem that integrates data about restaurants and movie theatres and places the information on a map[Barish et al.; 2000]. Moreover, Ariadne has been applied to integrate online electronic catalogues toprovide a more comprehensive virtual catalogue [Knoblock et al.; 2001].

The project OBSERVER (http://siul02.si.ehu.es/~jirgbdat/OBSERVER/) is developed by University ofPaís Vasco, MCC and the University of Georgia. Its aim is to integrate heterogeneous informationthrough domain ontologies [Mena et al.; 2001]. The OBSERVER model is being used as a part of aproject funded by the Government of the Basque Country, whose goal is to create and manage a globalinformation system of Eusko Ikaskuntza, an organisation that compiles and manages data about theBasque culture [Mena et al.; 98].

3.4.4 References[Guarino et al., 1999], Nicola Guarino, Claudio Masolo, Guido Vetere, "OntoSeek: Content-Based Access to the Web", IEEEIntelligent Systems, 1999.[Deborah L. McGuinness, 1998], Deborah L. McGuinness, "Ontological Issues for Knowledge-Enhanced Search",[Stuart Aitken 2000] Stuart Aitken, Sandy Reid,"Evaluation of an Ontology-Based Information retrieval Tool" , ECAI'00,Applications of Ontologies and Problem-Solving Methods.[Fensel et al., 2000], Dieter Fensel, Stefan Decker, Michael Erdmann, Rudi Studer, "Ontobroker: Ontology based Access toDistributed and Semi-Structured Information"[Qi Li et al 2001], Qi Li, Philip Shilane, Natalya Fridman Noy, M.A. Musen “Ontology Acquisition from on-line KnowledgeSources”, AMIA Inc. pp. 497 - 501[L. Crow 2001], L. Crow, N. Shadbolt “Extracting focused knowledge from the semantic web” Int. J. Human-Computer Studies54, pp 155-184[S. A. McIlraith 2001], S.A. McIlraith, T.C Son, H. Zeng “Semantic Web Services”. IEEE Intelligent Systems, pp. 46-53[Sure et al., 2002], York Sure, Michael Erdmann, Jürgen Angele, Steffen Staab, Rudi Studer, Dirk Wenke. OntoEdit: Collaborativeontology development for the semantic web. InProceedings of the ISWC 2002, June 9-12 2002, Sardinia, Italia., 2002.

[Arens et al.; 1996] Arens Y., Knoblock C.A., Shen W.M., “Query reformulation for dynamic information integration ”, Journalof Intelligent Information Systems, Special Issue on Intelligent Information Integration, 6(2/3). pp.99-130, 1996.

[Barish et al.; 2000] Barish G., Knoblock C.A., Chen Y.S, Minton S., Philpot A., Shahabi C., “The theaterloc virtual application”,In Proceedings of Twelfth Annual Conference on Innovative Applications of Artificial Intelligence (IAAI-2000), Austin, Texas,2000.

[Knoblock et al.; 2001] Knoblock A.C., Minton S., Ambite J.L., Ashish N., Muslea I., Philpot A.G., Tejada S., “The ariadneapproach to web-based information integration”, To appear in theInternational the Journal on Cooperative Information Systems(IJCIS) Special Issue on Intelligent Information Agents: Theory and Applications, 10(1/2), pp. 145-169, 2001.

[Mena et al.; 2001] Mena E., Illarramendi I., "Ontology-Based Query Processing for Global Information Systems", KluwerAcademic Publishers, ISBN 0-7923-7375-8, pp. 215, 2001. June 2001.

[Mena, 1998] Mena E., "OBSERVER: An Approach for Query Processing in Global Information Systems based onInteroperation across Pre-existing Ontologies", PhD. thesis, University of Zaragoza, November 1998.

SARI : http://www.ercim.org/publication/Ercim_News/enw35/taveter.html

MCM: http://ai.bpa.arizona.edu/go/intranet/papers/Customizable-00.htm


39/100

3.5 Portals and Web communities

According to [Bayton, 2002], portal technology is the latest must-have tool being snapped up by forward-looking for B2B (business to business), B2C (business to customers) and B2E (business to employee)communications. However, before companies dedicate considerable portion of a budget to portalimplementation and development, it is strongly encourages a back-to-basics approach in whichestablishing a solid strategy, setting corporate objectives and understanding the importance of informationretrieval are essential ingredients.It is far too easy for the portal generation to succumb to the same pitfalls as the misjudged e-commercegeneration before it. Like an e-commerce strategy, a portal strategy can have a dramatic impact on thebusiness of a company. Starting point for a company is to try to answer itself why it wants a corporateportal ? A recent Butler report [Butler,2001] highlighted three main reasons for building and deploying aportal: distributing information more effectively; encouraging collaborative working; and managingcontent and information. These reasons amount to improved individual productivity. Thus, the keyconsideration for choosing any portal therefore has to be whether or not it will dramatically improve thecompany’s efficiency levels.The current economic climate dictates that such a consideration must be supported by a solid return oninvestment (ROI). However, it is not only a question of expenditures estimation. It is above all a questionof demonstrating how the real portal’s impact on individual productivity can properly determine the ROIfor the company. General problem for a company is then to answer how can a portal improve anindividual’s productivity? Most of the time, portal consultants use to demonstrate the key features of aportal in one or two ways. The first one is through individual personalisation, which is one of the leadingarguments behind installing a corporate portal. Such argument conveys the idea that each individualworker may have its own unique window on the company, giving to him instant access to the vitalinformation required to perform more productively. The second one is far more compelling, namely asingle point of access to information and an information search facility. In essence, a portal’s key featurehas to be its ability to find what you want, when you want it. In that sense, information retrieval (IR) andhow effectively and efficiently it is performed, is the key to boost individual’s productivity. Providingevery employee with instant access to every relevant piece of information held within it will enablegreater efficiency, allowing faster and more accurate decision making, stronger collaborative working andhigher level of cross-company knowledge sharing. As a consequence to this, how the company willstructure its search and retrieval feature will be vital to have a successful portal.


The widely-agreed core idea of the Semantic Web is the delivery of data on a semantic basis. Intuitivelythe delivery of semantically processable data should help with establishing a higher quality ofcommunication between the information provider and the consumer. The vision of the Semantic Web isclosely related to ontologies as a sound semantic basis that is used to define the meaning of terms andhence to support intelligent providing and access to information for Web communities.

Navigating through Web portals which are based on a topic thesaurus, likehttp://dmoz.orgorhttp://www.yahoo.com4, is more or less equivalent to browsing a static hierarchy. Those with a richersemantic model, such as KA2Portal [AIF00] (http://ka2portal.aifb.uni-karlsruhe.de), offer simplenavigation through a class hierarchy.

4 In contrast to the SEAL approach Yahoo only utilizes a very light-weight ontology that solely consists of categories

arranged in a hierarchical manner. Yahoo offers keyword search (local to a selected topic or global) in addition to

hierarchical navigation, but is only able to retrieve complete documents,i.e. it is not able to answer queries

concerning the contents of documents, not to mention to combine facts being found in different documents or to

include facts that could be derived through ontological axioms.


40/100

The Ontobroker project [Ontobroker 1999] lays the technological foundations for the AIFB portal. Ontop of Ontobroker the portal has been built and organizational structures for developing and maintaining ithave been established. The approach closest to Ontobroker isSHOE [SHOE 2000]. In SHOE, HTMLpages are annotated via ontologies to support information retrieval based on semantic information.Besides the use of ontologies and the annotation of web pages the underlying philosophy of both systemsdiffers significantly: SHOE uses description logic as its basic representation formalism, but it offers onlyvery limited inferencing capabilities. Ontobroker relies on Frame-Logic and supports complexinferencing for query answering. Furthermore, the SHOE search tool does not provide means for asemantic ranking of query results. A more detailed comparison to other portal approaches and underlyingmethods may be found in [Staab S. et al. 2000].

A richer semantic model, such asSEAL Portal [Stojanovic et al. 2001] or C-Web [Saglio et al.2002], offer navigation through a class hierarchy for dynamic exploration.Such framework should help users to navigate through portals leading into very large resource spaces,and to find quickly many resources but only those of interest for them.

3.5.2 Use Cases and Needs

The overall architecture and environment of a classical portal for a Web community is well illustrated by[Stojanovic et al. 2001] (figure 3.8). Thebackboneof the system consists of theknowledge warehouse i.e.the ontology and knowledge base, and theOntobrokersystem,i.e. the principal inferencing mechanism.The latter functions as a kind of middleware run-time system, possibly mediating between differentinformation sources when the environment becomes more complex than it is now.At the front end one may distinguish between three types ofagents: software agents, community usersandgeneral users. All three of them communicate with the system through theweb server.The three different types of agents correspond to three primary modes of interaction with the system.

First , remote applications (e.g. software agents) may process information stored in the portal. For thispurpose, theRDF generatorpresents RDF facts through the web server. Software agents withRDFcrawlersmay collect the facts and, thus, have direct access to semantic knowledge stored at the web site.

Second, Community users and general users can access information contained at the web site. Two formsof accessing are supported: navigating through the portal by exploiting hyperlink structure of documentsand searching for information by posting queries. The hyperlink structure is partially given by the portalbuilder, but it may be extended with the help of thenavigationmodule. The navigation module exploitsinferencing capabilities of the inference engine in order to construct conceptual hyperlink structures.Searching and querying is performed via thequery module. In addition, the user can personalize thesearch interface using thesemantic personalizationpreprocessing module and/or rank retrieved resultsaccording to semantic similarity (done by the post-processing module forsemantic ranking). Queries alsotake advantage of the Ontobroker inferencing capabilities.

Third , only community users can provide data. Typical information they contribute includes personaldata, information about research areas, publications, activities and other research information. For eachtype of information they may contribute there is (at least) one concept in the ontology. By retrieving partsof the ontology, thetemplatemodule may semi-automatically produce suitable HTML forms for datainput. The community users fill in these forms and the template modules stores the data in the knowledgewarehouse.


41/100

Figure 3.8 - SEAL Overall architecture and Use Case of Classical Portal for Web Community

3.5.3 Representative Applications

3.5.3.1 SEAL - A Framework for developing Semantic Portals

The topic of this section is SEAL (SEmantic PortAL), a framework for managing community web sitesand web portals on an ontology basis. The ontology supports queries to multiple sources (a task alsosupported by semi-structured data models [Fernandez, 2000]), but beyond that it also includes theintensive use of the schema information itself allowing for automatic generation of navigational viewsand mixed ontology and content-based presentation. The core idea of SEAL is that Semantic Portals for acommunity of users that contributeandconsume information [Staab, 2000] require web site managementand web information integration. In order to reduce engineering and maintenance efforts SEAL usesontology for semantic integration of existing data sources as well as for web site management andpresentation to the outside world. SEAL exploits the ontology to offer mechanisms for acquiring,structuring and sharing information between human and/or machine agents. Thus, SEAL combines theadvantages of the two worlds briefly sketched above.

The SEAL conceptual architecture (cf. figure 3.9); details to be explained in subsequent sections) depictsthe general scheme. Approaches for web site management emphasize on the upper part of the figure andapproaches for web information integration focus on the lower part while SEAL combines both withontology as the knot in the middle.


42/100

Figure 3.9 - SEAL conceptual architecture

For more information on SEAL history and technical aspects, reader is invited to consult annex 1.

3.5.3.1.1SEAL technical Architecture

Figure 3.10 - KAON framework

The technical architecture of SEAL is derived from the architecture of KAON, the Karlsruhe SemanticWeb and Ontology Infrastructure5, whose components provide the required functionalities described inthe previous sections. The architecture of KAON is depicted in figure 3.10.KAON components canroughly be grouped into three layers.

5 http://kaon.semanticweb.org


43/100

The data and remote services layerrepresents optional external services, which can be used in theupper layers, e.g. reasoning services for inferencing and querying, or connectors to the Edutella Peer-To-Peer network, and alternative storage mechanisms for the data in the previously mentioned warehouse.

The middleware layer provides a high-level API for manipulating ontologies and associated data andhides the actual manner of storage and communication from all clients. Thus clients cannot distinguishbetween working on the local file system (provided by the RDF API) or working on a multi-user awareserver which stores data in a relational database. The middleware also provides interfaces to QEL, thequery language used within the Edutella network, which is not only used to communicate queries withinthe peer-to-peer network but also used to query the warehouse.

The application and services layergroups applications that use services from the underlying layers.Currently these are one hand, stand-alone desktop applications built using the Ont-O-Mat applicationframework or portals built using the KAON portal maker, which provides the features discussed in theprevious section. Ont-O-Mat applications are built as plug-ins that are hosted by the Ont-O-Matapplication framework. This approach guarantees maximum application interoperability within Ont-O-Mat.

Finally, core to KAON is the domain ontology itself, which is represented in RDF Schema [W3C, 1999] -the data model at hand for representing ontologies in the Semantic Web. It provides basic class andproperty hierarchies and relations between classes and objects. Historically SEAL leverages the mappingof RDF Schema model to F-Logic [Kifer, 1995] introduced in [Decker, 1998 7] to provide views (in formof logical axioms) and a query mechanism. This allows us to rely on the reasoning services offered byOntoBroker [Decker, 1999] or SiLRi [Decker, 1998 7].

3.5.3.1.2 ConclusionIn addition to ongoing work to integrate Peer-to-Peer functions for accessing information on the Web,two topics are currently under investigation: first, the view concept that is implemented by the KAONframework does not support updates in general. Currently, only the simplistic input views provide meansfor updating the warehouse. Clearly, Web site users do expect to be able to update the site content. Asecond topic that needs further improvement is the handling of ontologies. Just offering a single,centralized ontology for all Web site users does not meet the requirements for heterogeneous user groups.Therefore, methods and tools are under development that support the handling and aligning of multipleontologies.

The SEAL framework as well as the KAON infrastructure can be seen as steps for realizing the idea ofthe Semantic Web. Obviously, further steps are needed to transfer these approaches into practice.

For more information about SEAL and KAON, the reader is invited to consult annex 1 and 2.

3.5.3.2 C-Web: Community Webs

C_WEB is a collaborative effort, associating several research organizations, IT companies and "advancedusers", aiming at designing a generic platform based on open standards and distributed as open source,and the related methodology and know-how, to support community-webs. C-Web consortium intends tovalidate both the software platform and the methodology through experiments carried-out with severaluser communities.


44/100

3.5.4 References[Bayton, 2002], M. Bayton (Director of strategic alliances – Convera), “Enter the mystical portal – Examining information

retrieval and the search cycle”, In. Content Management Focus, Vol.1, n°3, December 2001/January 2002.

[Butler et al, 2001], M. Butler, I. Charlesworth, G. Cooper, M. Davis, K. Gibbs, J. Halé, T. Jennings, D. Lightfoot, I. Lynch, M.

Thompson, “Enterprise Portals Report” , Butler report, July 2001.http://www.butlergroup.com/reports/eportals/

[Campbell, 2000], I. Campbell (Vice President Research – Nucleus), “Maximising return on investment from enterprise portal”,

Nucleus Research Inc. Report, March 2000.

URL: http://www.NucleusResearch.com

[CIO2002], CIO Custom publishing, “When Business communication is critical”, Advertising supplement, February 2002.

[Stojanovic et al., 2001], N. Stojanovic, A. Maedche, S. Staab, R. Studer, and Y. Sure. "SEAL— A Framework for Developing

SEmantic PortALs" . In K-Cap 2001 - First International Conference on Knowledge Capture, Oct. 21-23, 2001, Victoria, B.C.,

Canada, 2001. to appear.

URL : http://ontobroker.semanticweb.org/ontos/aifb.html

[Saglio et al. 2002], J. M. Saglio, Tuang Anh Ta, "A Framework for Dynamic Exploration in Semantic Portals", KR2002 to appear

C-Web : Community Webs

URL: http://cweb.inria.fr/

[Staab S. et al. 2000] S. Staab, J. Angele, S. Decker, A. Hotho, A. Maedche, H-P. Schnurr, R. Studer, and Y. Sure. "AI for the web

— ontology-based community web portals". In AAAI 2000/IAAI 2000 - Proceedings of the 17th National Conference on Artificial

Intelligence and 12th Innovative Applications of Artificial Intelligence Conference. AAAI Press/MIT Press, 2000.

[Microsoft, 2001], Microsoft Portal team, “Introducing Microsoft SharePoint™ Portal Server 2001”, Microsoft white paper,

May 2001,http://www.microsoft.com/sharepoint/techinfo/planning/SPSIntro.doc.

[Ontobroker, 1999] S. Decker, M. Erdmann, D. Fensel, and R. Studer. "Ontobroker: Ontology Based Access to Distributed and

Semi-Structured Information". In R. Meersman et al., editors,Database Semantics: Semantic Issues in Multimedia Systems, pages

351–369. Kluwer Academic Publisher, 1999.

[Plumtree, 2001] Plumtree Inc., “A Framework for Assessing Return on Investment for a Corporate Portal Deployment”,

Company Report :http://www.plumtree.com

[SHOE 2000] J. Heflin and J. Hendler. "Searching the web with SHOE". In Artificial Intelligence for Web Search. Papers from

the AAAI Workshop. WS-00-01, pages 35–40. AAAI Press, 2000.

[Fernandez, 2000] M. F. Fernandez, D. Florescu, A. Y. Levy, and D. Suciu.Declarative specification of web sites with Strudel.VLDB Journal, 9(1):38–55, 2000.

[Staab, 2000] S. Staab, J. Angele, S. Decker, M. Erdmann, A. Hotho, A. Maedche, H.-P. Schnurr, R. Studer, and Y. Sure.Semanticcommunity web portals. In WWW9 / Computer Networks (Special Issue: WWW9 - Proceedings of the 9th International WorldWide Web Conference, Amsterdam, The Netherlands, May, 15-19, 2000), volume 33, pages 473–491. Elsevier, 2000.

[W3C, 1999] W3C.RDF Schema Specification. http://www.w3.org/TR/PR-rdf-schema/, 1999.

[Kifer, 1995] M. Kifer, G. Lausen, and J. Wu.Logical foundations of object-oriented and frame-based languages. Journal of theACM, 42:741–843, 1995.

[Decker, 1998] S. Decker, D. Brickley, J. Saarela, and J. Angele.A query and inference service for RDF. In QL98 - QueryLanguages Workshop, December 1998.

[Decker, 1999] S. Decker, M. Erdmann, D. Fensel, and R. Studer.Ontobroker: Ontology based access to distributed and semi-structured information . In R. Meersman et al., editors,Database Semantics: Semantic Issues in Multimedia Systems, pages 351–369. Kluwer Academic Publisher, 1999.


45/100

4 Evaluation of Ontology Based Applications

4.1 Why evaluate ontology-based applications?

In the increasingly competitive knowledge-intensive economy, the search for competitive advantage andthe necessity to optimally allocate scarce resources creates a pressing need for evaluating theROI as wellas the contribution tovalue creation of applications. This is especially true of IT and knowledge-intensive applications for which this contribution is not always clear and often difficult to justify.

The semantic-web is a relatively new concept whose stated vision is generally accepted by industry, eventhough there are as yet few successful high-profile cases that support the theory and demonstrate tangibleresults. However, in order to avoid mistakes that have been made in the past with other leading edge, highprofile technologies such as AI, it is essential that the semantic web community devise evaluationmethodologies that they will be able to apply from the initial stages of their effort. There are two mainreasons why this is useful:

ÿ� Management of expectations: As with all new things it is always wise to“promise less and delivermore” . Evaluation of ontology-based systems will indicate the scope and limitations of suchapplications thus allowing developers to understand the capabilities of the technology anddeployment time frames and end users to be educated about system potential.

ÿ� Guidanceof the development and implementation process:evaluation of applications is useful todevelopers since it provides information and feedback that can be used in their further development.

It is an important industrial need to deliver high-quality Ontology-based systems. Evaluation is requiredto ensure this high quality and to guide the development and maintenance. Problems with existingevaluation studies [Menzies, 1998] show that there is acrucial need for a systematic methodologythathelps in conducting"good measurements".

To date, the knowledge technologies research community has studied a number of issues relating to theaccess, management and reuse of information/knowledge predominantly in qualitative ways.Quantitative analyses are relatively rare, although [Cohen, 1999], [Gomez-Perez 1999] from theontology side, [Deborah L. McGuinness 1999] from the IR side and the use of software engineeringapproaches to evaluation [Menzies, 1998] and [Nick, 1999] are first efforts in this direction.


46/100

4.2 Criteria or metrics for the evaluation of ontologies

Ontologies aim at capturing domain knowledge in a generic way and provide a commonly agreedunderstanding of a domain, which may be reused and shared across applications and groups[Chandrasekaran et al.; 99]. Ontologies provide a common vocabulary of an area and define -withdifferent levels of formality- the meaning of the terms and the relations between them. Ontologies areusually organized in taxonomies and typically contain modeling primitives such as classes, relations,functions, axioms and instances [Gruber, 93].

Based on Gruber's definition [Gruber, 93]: "An ontology is an explicit specification of aconceptualisation", many definitions of ontologies have been proposed. In 1995, Guarino and Giaretta[Guarino et al., 95] collected seven definitions and provided corresponding syntactic and semanticinterpretations. In 1997, Borst [Borst, 97] slightly modified Gruber's definition saying that: ``Ontologiesare defined as a formal specification of a shared conceptualisation''.

Studer [Studer et al., 98] has explained these two definitions as follows: "Conceptualisation refers to anabstract model of some phenomenon in the world by having identified the relevant concepts of thatphenomenon. Explicit means that the type of concepts used, and the constraints on their use are explicitlydefined. Formal refers to the fact that the ontology should be machine-readable. Shared reflects the notionthat an ontology captures consensual knowledge, that is, it is not private to some individual, but acceptedby a group.''

User communities are rarely homogeneous. For a company Intranet, for example, different departments inthe company will:

ÿ� Create information resources using specific terms and

ÿ� Ask questions using different terms.

Different levels of worker will also use different sets of terms. The terms used in management reports willbe different from that used in documents made available to the public. The terms used by domain expertsin e-mails will differ from those used by trainees during training courses. Any evaluation techniqueshould seek to measure the effectiveness of the ontology at many different levels, and not treat its use as asingle whole. It is important that each major sub-community’s needs be evaluated separately.Ranking the benefits of the system for communities will depend on the economic or service value of thesystem to each community.

When weighting terms in ontology based on their use it is important to weight them separately in terms oftheir occurrence within information resources and their occurrence within search requests. Ontologiesshould be evaluated against representative selections of information resources. For example, for a virtualnetwork the ontology must be tested separately against the information being provided by each partner,with term weighting being recorded on a partner-by-partner basis rather than on a system-wide basis.Term usage in queries can also be expected to change on a partner-by-partner basis.

The main purpose of any ontology is to resolve mismatches between the terms used in seekinginformation and those used in providing information. Therefore the effectiveness of any ontology isdetermined by its ability to match a search term with a term in an information resource that has anequivalent meaning. This means that relationships between terms need to be weighted according to theaccuracy with which they match each other in terms of “replacability” within a question or informationsource (e.g. their cognitive adequacy). Terms also need to be weighted according to their specificity. Amore specific term should be more highly weighted than a less specific term. However, the requirementsof specificity are dependent on the user community. A system designed to be used by people outside ofthe community supplying the information is more likely to use generalized terms than terms used by the


47/100

experts providing the information. A system, designed to be used by domain experts, should give thehighest priority to the most specific terms. Metrics are required for measuring and recording these criteria.

Many different theoretical dimensions can be assessed for the evaluation of ontologies and for theircomparison:

ÿ� Usability,ÿ� Expressivity,ÿ� Accuracy,ÿ� Consistency,ÿ� Completeness,ÿ� Conciseness,ÿ� Expandability,ÿ� Sensitiveness,ÿ� Domain richness,ÿ� Cognitive adequacy.

In the following, we will try to: i) textually and graphically describes these metrics and then, ii) formalizetheir significance.

4.2.1 Generic Criteria

If ontologies are to be useful in industrial settings they should aim to satisfy a number of criteria thatreflect basic requirements of the target user communities.To date a number of researchers/organizations have published such criteria and in this section we willpresent a set of criteria that approaches in our view what might be a consensus view.

In evaluating ontologies it is useful to organize criteria into categories that will allow the interested partyto compare options in a consistent and meaningful manner. For a finer-grained evaluation, the categoriesthemselves can be further grouped according to the envisioned use of the ontology.

Generic CriteriaBased on the discussion of the previous sections we propose the following categories for evaluation:

ÿ� Modeling Capabilities: refers to the expressive richness of the underlying language/formalism,ÿ� Supporting tools: refers to infrastructure issues that make the use of the ontology viable,ÿ� Performance: a measure of the behavior in terms of computing resources required and kinds of

queries available,ÿ� Practicalities: deployment issues such as scope of use, external connectivity, support.

We present the criteria within these categories first in the context of the use of an ontology as arepresentation medium. We then consider each of these categories in the context of information retrieval.

Modeling Capabilitiesÿ� Does the language support the following?

o Multiple inheritance,o Value constraints,o Default values,o Conjunction, disjunction, negation,o Methods for value calculation,o Multi-valued properties.

ÿ� What primitive data types (e.g. numbers) does the language support?ÿ� Does the language support instances as well as classes?ÿ� If the language supports methods and/or constraints how rich is the formalism?ÿ� Does the language support multiple versions of data? Does the language support time

stamping of data?


48/100

Supporting toolsÿ� What tools exist for maintaining the ontology?ÿ� Are the tools themselves maintained?ÿ� What types of queries can be expressed?ÿ� Are there any limits in the size of the ontology, names, values etc.?ÿ� What kinds of reasoning mechanisms does the ontology support?

Performanceÿ� What queries can be expressed in the language ?ÿ� Parents/Children of Concept X,ÿ� Least common parent of concepts X and Y,ÿ� Closest common child of concepts X and Y,ÿ� Links between concepts X and Y,ÿ� Is there a language for expressing these queries?ÿ� What are the computing resources requirements of the language?

Practicalitiesÿ� Is the ontology being used in any real applications? How large?ÿ� How widespread is its use?ÿ� Is any industry-wide body backing it?ÿ� Links with other ontologies: Are there tools that allow the ontology to be linked with other

ontologies?ÿ� What support is available for the ontology formalism (manuals, tutorials, support team)?ÿ� Does a stable release (not to change within the next 6 months) exist?

4.2.2 UsabilityIn a general way, the usability of ontology can be defined as its adequacy for answering specific needs orrequests in a particular domain. In that way, it was already shown in the knowledge literature that it isoften difficult to simultaneously obtain high usability (i.e., adequacy for a specific use) and reusability(i.e., adequacy for several uses).

4.2.3 ExpressivityExpressivity determines what can be said by the ontology. An expressive ontology will contain a rich setof primitives or axioms that will allow a wide variety of knowledge to be formalized.

Figure 4.1 - The intended models of a logical language reflect its commitment to aconceptualisation. Ontology indirectly reflects this commitment (and the underlying

conceptualisation) by approximating this set of intended models.

Thus, an ontology with too little expressivity will provide too few reasoning opportunities to be of muchuse and may not provide any contribution over existing ontologies. Regularly, there is a differencebetween what we want to express with ontology and what is finally expressed by the ontology itself. This

ConceptualizationC

Language LCommitmentK=<C, �� >

OntologyOntology

ModelsM(L)

IntendedmodelsI K(L)

IntendedmodelsI K(L)


49/100

remark remains the conclusions of Guarino [8] dealing with the problem of ontologies and moreparticularly the fact they are approximate characterization of a domain.

According to Guarino, given a languageL with ontological commitmentK , ontology for L is a set ofaxioms designed in a way such that the set of its models approximates as best as possible the set ofintended models ofL according toK .Thus, expressivity can be formalized as the rendering ontology is able to provide both in consistency withthe intended model IK(L) and the conceptualization it commits to.

4.2.4 Accuracy

Ontologies characterize the conceptualization they are committing to, but with certain accuracy. In thatcontext, accuracy measures how close to the conceptualization is the corresponding ontology. Thus,depending on the way the ontology was built to characterize this conceptualization; different levels ofaccuracy may be obtained providing good or bad ontologies (Fig.7).

Figure 4.2 - Bad ontology vs. good ontology.

Bad ontologies will be much larger as the considered intended model IK(L) so that it may badly answerrequests on a concept or domain. On the opposite, good ontologies closely suite to a considered model, sothat they perfectly characterize the conceptualization they commit.

The next described following criteria have been identified by Asun Gómez-Pérez in [Gómez-Pérez,2002].

4.2.5 Consistency

Consistency refers to whether it is possible to obtain contradictory conclusions from valid inputdefinitions. A given definition is consistent if and only if the individual definition is consistent and nocontradictory sentences can be inferred using other definitions and axioms.

4.2.6 Completeness

Incompleteness is a fundamental problem in ontologies. In fact, we cannot prove either the completenessof an ontology or the completeness of its definitions, but we can prove both the incompleteness of anindividual definition, and thus deduce the incompleteness of an ontology, and the incompleteness of anontology if at least one definition is missing with respect to the established reference framework.

� � ��úù�ö ��

�úù�ö ��

�úù�ö ��

Bad ontologyBad ontology


50/100

4.2.7Conciseness

An ontology is concise if it does not store any unnecessary or useless definitions, if explicit redundanciesdo not exist between definitions, and redundancies cannot be inferred using other definitions and axioms.

4.2.8 Expandability

Expandability refers to the effort required in adding new definitions to an ontology and more knowledgeto its definitions, without altering the set of well-defined properties that are already guaranteed.

4.2.9Sensitiveness

Sensitiveness relates to how small changes in a definition alter the set of well-defined properties that arealready guaranteed.


51/100

4.3 Evaluation in Information Retrieval

4.3.1 Reference evaluation methodology

Evaluations in IR has been launched formally in early 90's through the NIST (National Institute ofStandards and Technology) and DARPA (Defense Advanced Research Projects Agency) and largelyreported through the annual TREC conference series [TREC web site], which is considered asthereference for test collection and benchmark tasks6. The goal of the conference series is to encourageresearch in information retrieval from large text applications by providing a large test collection, uniformscoring procedures, and a forum for organisations interested in comparing their results. The first TRECconference was held in November 1992.

The TRECtest collectionis composed of:ÿ� The document collection (about 8 gigabytes of various text document)ÿ� Example Information Requests ("Topics")ÿ� Relevant documents for each topics (as those expected from each Topic)

ÿ� Set oftasksÿ� Basic information requestÿ� Filtering (on a dynamically changing document corpus)ÿ� Interactive (how the dialogue improve the relevance of the documents return)ÿ� Natural Language requests (expected improvements brought by associated NLP)ÿ� Cross-Languages request (on multilingual corpus)ÿ� High precision (Five minutes to retrieve ten documents that match a previously unknown query)ÿ� Spoken document retrieval (e.g. broadcast news)ÿ� Very large corpus (8 millions documents)

ÿ� Evaluation measuresat the TREC conferences7

ÿ� Statistics summary for each task (number of documents for all topics, …)ÿ� Recall-Precision averagesÿ� Document level averages (average precision computed at 5, 10, … relevant documents retrieved)ÿ� Average precision histogram (single R-precision for each topic vs average R-precision obtained

by all the IR systems participating in the evaluation)

6 Other test collections exist: CACM (Communication of the ACM papers), CIS ("cystic fibrosis" indexeddocuments from National Library of Medicine - MEDLINE), INSPEC (abstracts on electronics, computerand physics), CISI from ISI (Institute of Scientific Information).7 See Annex 4 for a full description of the measures


52/100

4.3.2 Reference evaluation criteria8

Retrieval evaluation benefits from a large experience since the early 80's, so a set relatively pertinentcriteria has been devised and defined formally to help in the evaluation of the retrieval performances ofdifferent retrieval techniques.

The authors usually distinguish between:ÿ� Query process in batch mode (i.e. the user submits a query and receives an answer back)ÿ� Query process in an interactive session (the user dialogues or cooperates with the system to get the

answers)When evaluating Ontology enhanced retrieval, both query processes are concerned with the Ontology andassociated reasoning as this is outlined in section 4.3.3.

Then, a set of distinct criteria is in use nowadays as follows

4.3.2.1 Recall and Precision

ÿ� Recall is the fraction of the relevant documents which has been retrieved

ÿ� Precision is the fraction of the retrieved documents which is relevant

Usually Recall and Precision are plotted in a graph P = f (R) according to a sorting parameter (e.g. rankposition in the displayed results)

8This part has been build with material from the chapter "Retrieval Evaluation" from: i) "Introduction to modern information

retrieval", G. Salton, M.J. McGill - McGraw-Hill publishing company 1983 ; ii) "Modern information retrieval", Ricardo Baeza-

Yates, Berthier Ribeiro-Neto - Addison Wesley publishers 1999.

Test Collection

Relevant

Documents [B] Retrieved

Documents [A]

Relevant

retrieved

documents

[R]

Precision = Card R / Card A

Recall = Card R / Card B

Precision

Recall

100%

100%


53/100

Average Precision / Recall measure

Usually, the retrieval technique is evaluated by running them on a set of Queries. So,the Precision vsRecall figuresare averaged over the number of Queries at a given Recall level.

That measure is extensively used as a standard evaluation strategy.

Variations on that measurehave been proposed and usedÿ� Average precision at document cutoff values(e.g. 10, 20, … first ranked documents), so

measuring the ranking algorithm performance.ÿ� Single value Precisionfigure gives the Precision value for each individual query, so allowing a

close-up analysis of the retrieval technique (e.g. anomalies, fine grain comparison)ÿ� Average precision at seen relevant document, so favoring the systems that shows relevant

documents quicker.ÿ� Precision histogramsshow the compared history of two retrieval techniques for each query. That

graph allows a quick visual comparison of the retrieval performances of two retrieval techniques.

4.3.2.2 Complementary metrics

Precisions versus Recall metrics suffer some critics, like the ability to measure precisely the Recall figurefor a very large, uncertain and very dynamic Documents collection (i.e. the Web). Also, as alreadymentioned, interactivity with the system and new display strategies are nowadays key aspects of theretrieval mode and so, the metrics should include measures that are not based on the hypothesis of a linearordering (i.e. ranking) of the pertinent retrieved documents.

Some complementary metrics have been proposed.ÿ� The Harmonic mean which combines Precisionÿþý and Recallÿ�ý measures on the range [0, 1]

giving a value 0 when no relevant documents have been retrieved and 1 when all ranked documentsare relevant.�

�ÿ�ý�� ÿ�ý�� þÿ�ý�

ÿ� E-Measure, which combines Precisionÿþý and Recallÿ�ý measures that, allows the user to favoreither Precision or Recall in the evaluation.

�

�ÿ�ý = ��ÿ�� ÿ�ý��þÿ�ý�ý

ÿ� Coverage ratiomeasures the ratio of the retrieved documents over the relevant documents known tothe user.

ÿ� Novelty measures the ratio of the relevant documents which were known to the userÿ� Relative recall measures the ratio between the number of relevant retrieved documents and those

expected by the user.ÿ� Recall effort measures the ratio between the number of relevant documents the user expected to find

and the number of examined retrieved documents.


54/100

4.3.3 Specific Ontology enhanced criteriaOntologies are more than simple lists of concepts. Typically an ontology will be a taxonomy of conceptsarranged in anis-a hierarchy and often other relations (such as the ‘part-of’ relation) will also berepresented. Given that these relationships are the main value added by an ontology, the evaluation of anontology-based IR system must also assess the benefits derived from this embedded knowledge. Forexample, using theis-a and/orpart-of relations, a query could be augmented (broadened or narrowed) tocreate a more precise context for the IR step. Other less common relationships such as functionalcausality, spatial proximity and biological pathways could also be used and their effect on systemperformance assessed.

The criteria discussed above are classical IR evaluation criteria in that they are applicable irrespective oftheir intended use of ontology. Using ontology-based IR systems should bring the following expectedimprovements:

Precision and recall of the system

Precision should be closer to 100% earlier in the relevance ranking of the retrieved documents

Recall and Precision are the two most basic criteria for evaluating IR systems. The exact performancelevels will depend on the requirements of the end user and the criticality of the application. Current state-of-the-art can often achieve performance in excess of the 90% mark especially in the original training set(the data set on which the design of the IR system is based) and of the order of 85-95% in other data sets.Furthermore if the ontology-based solution is to replace an existing system (presumable keyword based)it is also reasonable to expect that the former will be better than that of the latter. Table 2 belowsummarizes these metrics based on recall and precision.

Original Data set New Data set

Relative Performance ONT-Based > KWD-Based ONT-Based > KWD-Based

Absolute performanceOf Ontology-based IR

Recall > 95%Precision > 90 %

Recall > 85%Precision > 80 %

Table 4.1 - Evaluation Method and Metrics for ontology-based versus keyword-based IR usingrecall and precision as criteria

Presentation of the context of the Query formulation and answerOntologies are primarily used for site map organization and insupport of browsing. So, viewed as aconceptualization of a domain, ontologies should provide some support to present the currentquery/answer context.

Precision

Recall

100%

100%


55/100

Augmenting of the user’s queryOntologies viewed as a more or less formal semantic network can provide - behind the scene - contextunderstanding and finallyquery expansionin searching. Also Query may be transformed into aprecisequery language- thanks to the formal KR that may be used - with well-defined semantics. This shouldalso lead toparametric search. It should facilitate use ofadvanced searchfunctions without requiringknowledge of a search language.

Guidance in query building and refinementsQuery formulation should be guided by the underlying ontologies (e.g. ontology-driven dialogue,ontology-driven form generation) andreformulation for "unsuccessful" queries.

Multi-modal query formulationDoes the system allow for different kinds ofcombined search modese.g. specific search, browsing,discovery?

Ranking of the retrieved documentsTherelevance rankingof the retrieved documents might be improved by the underlyingsemantics.Also, merging, abstracting and pruning of the retrieved contents could be done in more pertinent way.

What kinds of relevance criteria are being used on the query results? In the case of web pages is it numberof hits, number of links or some other measure?

Explaining the results

Does the system provide for explanations on Query answers or better on the no-answer cases? Ranking ofthe retrieved documents is a relatively poor and silent criterion on that respect. Answers that are returnedare not just explicit answers but also implicit ones. Does the system indicate possible contradictions ormultiple versions of the same information? Does the system allow for any kind of trend analysis ordetection in the query results?

Response timeHow timely is the information returned?

Data NoiseMany applications have to deal with source data that is far from perfect. Typical problems includemisspellings, synonyms, abbreviations and non-standard delimiters, which might degrade systemperformance. The issue of noise appears often in applications involving data collection (e.g. in faultassessment of engineering systems, in clinical trials and in customer support centers) where usually themost relevant information is entered in shorthand, not following typical grammar or syntax and usuallyinvolving a lot of different versions of domain specific terms. Ontology-based IR systems could in theorydeal with such noise (by providing a context which might eliminate the noise) and should therefore beassessed against this criterion as well.

RobustnessThe robustness of concept extraction is a factor in the assessment of the ontology, hence there is a need toassess adequacy and comparative performance on both the data set for which the system was constructed(and can be expected to perform well on) and on new data sets. Table 4.1 above illustrates this.

Larger coverageThe Ontologies providing a moreorganised and formalisedrepresentation of domains could ensure alarger coverage of the resources. Here we can envision that indexing be done directly on Metadata andthat inter-working at the semantic levelaugments the breadthof the covered resources.


56/100

Total ROIAs ontologies represent a significant effort in their creation, sharing and maintenance, the evaluationmethod should also account for the total investment in terms of man-hours versus the gains in precisionthat these methods are expected to yield in relation to traditional approaches (e.g. keyword-based IR).While simple applications with relatively small domain specific ontologies (of the order of a few hundredconcepts) may be easily implemented with 3-5 person-days, the creation and especially the maintenanceand re-use of larger organization (or industry) wide ontologies is a much more complex affair. To datethere is no study that assesses ontology-based IR systems in large applications taking total ROI intoconsideration.


57/100

4.4 Enterprise Portals and Knowledge Management

4.4.1 Reference evaluation methodologyRegarding actual enterprise portals and knowledge management literature, it appears that no referenceevaluation methodology can be identified yet. Such a situation may be explained by the two non-exhaustive following reasons:

ÿ� Evaluation methodology is most of the time strongly inter-linked to companies or organisationsbusiness plan, strategy or business objectives, so that i) it is company dependant, and ii) it mustbe kept confidential.

ÿ� Evaluation methodology could be considered as the end result of a process of thinking, whichusually starts from identification of pertinent criteria and metrics, in consistency with some pre-defined business objectives, and then, goes through the use of these criteria, in adequacy withsome clusters of results the company is looking for. As a consequence to this, a set of differentevaluation methodologies may appear depending on the kind of results the company isexpecting.

Regarding these two reasons, no reference and generic evaluation methodology is known and could bepresented yet.

4.4.2 Reference evaluation criteriaMeasuring and maximizing return on investment (ROI) from an enterprise portal9

Some keys to success in purchasing an enterprise portal for a company have been identified [Campbell,2000].

4.4.2.1 Identify a business objective, not an ideal portal

This very first step can be resume answering the following questions: what kind of portal will support thecompany’s objectives? And, what value can it bring?Here, we resume in table 4.2 the potential returns the company may have depending on the type of portalconsidered.

Employee-facing (B2E) Partner-facing (B2B) Customer-facing (B2C)

Reduced administrative overheadShortened sales cycle

Reduced communication costsReduced training and HR costs

Reduced administrative overheadReduced communication costs

Reduced cost of salesReduced accounts receivable

Reduced product rework expenses

Reduced cost of salesReduced administrative

overheadReduced communication costs

Table 4.2 - Potential returns per type of enterprise portal

Enterprise portals are mostly defined to answer some different needs in terms of information andinformation exchanges. Thus, a second key success for purchasing a portal will be to precisely identifythe fivefunctionality factors that may be considered:

Factor 1 - Application data integration:o How many applications / information resources will need to be accessed through the portal?o Can they be accessed through plug-ins or is development required?o How well/rapidly do potential vendors support integration?

9 [Campbell 2000]


58/100

Factor 2 - Searching and indexing:o How much does the user know about the data/applications they will access? (Key point)o How does the portal categorize/organize information?o Do the search capabilities meet the needs and expertise of the user?

Considering this crucial portal’s issue drives companies to make some technological choices in termsof knowledge management, taxonomy or ontology – based, and managing or not natural or/ andforeign languages.

Factor 3 - Personal productivity support:o Is employee productivity support an objective of the portal?o What technologies are employees using to support productivity today?o What agents, devices, and tools are available/appropriate?o What training is needed? (key point).

Factor 4 - Management and collaboration:o Are accelerated product development or sales cycles, or improved crisis management, an

objective of the portal?o What technologies are currently used for tactical collaboration and task management?o What technologies are available in a portal? (ontologies, taxonomies, etc.),o What training will be needed?

Factor 5 - Content provision:o Are reduced research or library costs objectives of the portal?o What external information sources do portal users need?

To measure the potential return of a portal, the aforementioned factors must be considered, looking forbreadth, repeatability, cost, collaboration and knowledge management.

In that sense, a report from BackWeb [CIO2002], company aiming at helping other companies tomaximize their content investment by prioritizing, delivering and promoting the usage of criticalinformation to customers (B2C), suppliers (B2B), partners and employees (B2E) across the enterprise, thereturn on investment for enterprise portal can be defined as following :Portal ROI (return on investment)= Content + Content Usage’s.

4.4.2.2 Looking for breadth and repeatability

Regarding the five previous factors, measurement of the potential return may be resume in table 4.3.

Factor 1Breadth

Factor 2Repeatability

Factor 3Cost

Factor 4Collaboration

Factor 5Knowledge

Does it impact a lot

of people or only a

few?

Will the application

be used frequently or

infrequently?

Is this a costly task

or relatively

inexpensive?

Does this task

involve collaboration

among groups?

Will this task involve

management of

key information?

The greater the

breadth of the

application, the

higher the potential

return

The greater the

repeatability of the

application, the


return

The greater thecostof the task, or the

greater the benefit,

the higher the

potential return

The greater the

collaborationcomponent of the

task, the higher the

potential return

The greater the use

of knowledgemanagementthe


return.

Table 4.3 - Factors for measuring the potential return of a portal for a company

Based-on these factor or criteria’s, it becomes possible to provide scoring on concrete case studies. Hereare some examples highlighting or not some potential benefits and returns for a company interested inbuilding an enterprise portal.


59/100

Example 1Use a portal to create a document-

centric workflow to change your

health care provider

Example 2Develop a customer-facing portal to

deliver accurate sales pricing and

product documents

Example 3Allow partners to access production

information to manage supply

delivery

ÿþýþüûüúùøùû÷�� ÿÿÿÿÿþÿÿÿÿÿÿÿÿýÿÿÿÿÿÿÿÿüÿÿÿÿÿÿÿÿûÿ

��þü�û��ÿÿÿÿÿþÿÿÿÿÿÿÿÿýÿÿÿÿÿÿÿÿüÿÿÿÿÿÿÿÿ � ÿ

� � �û��þ�þ�ùû�� ÿÿÿÿÿþÿÿÿÿÿÿÿÿýÿÿÿÿÿÿÿÿüÿÿÿÿÿÿÿÿûÿ

� �øøüú��üûù�� ÿÿÿÿÿþÿÿÿÿÿÿÿÿýÿÿÿÿÿÿÿÿüÿÿÿÿÿÿÿÿûÿ

�� øþ��þ� �ü�ü�þ�þ�û��ÿÿÿÿÿ � ÿÿÿÿÿÿÿÿýÿÿÿÿÿÿÿÿüÿÿÿÿÿÿÿÿûÿ

ÿþýþüûüúùøùû÷��ÿÿÿÿÿþÿÿÿÿÿÿÿÿýÿÿÿÿÿÿÿÿüÿÿÿÿÿÿÿÿ � ÿ

��þü�û��ÿÿÿÿÿþÿÿÿÿÿÿÿÿýÿÿÿÿÿÿÿÿ � ÿÿÿÿÿÿÿÿûÿ

� � �û��þ�þ�ùû��ÿÿÿÿÿþÿÿÿÿÿÿÿÿýÿÿÿÿÿÿÿÿ � ÿÿÿÿÿÿÿÿûÿ

� �øøüú��üûù��ÿÿÿÿÿ � ÿÿÿÿÿÿÿÿýÿÿÿÿÿÿÿÿüÿÿÿÿÿÿÿÿûÿ

�� øþ��þ� �ü�ü�þ�þ�û��ÿÿÿÿÿþÿÿÿÿÿÿÿÿýÿÿÿÿÿÿÿÿüÿÿÿÿÿÿÿÿ � ÿ

ÿþýþüûüúùøùû÷��ÿÿÿÿÿþÿÿÿÿÿÿÿÿýÿÿÿÿÿÿÿÿüÿÿÿÿÿÿÿÿ � ÿ

��þü�û��ÿÿÿÿÿþÿÿÿÿÿÿÿÿýÿÿÿÿÿÿÿÿ � ÿÿÿÿÿÿÿÿûÿ

� � �û��þ�þ�ùû��ÿÿÿÿÿþÿÿÿÿÿÿÿÿýÿÿÿÿÿÿÿÿüÿÿÿÿÿÿÿÿ � ÿ

� �øøüú��üûù�� ÿÿÿÿÿþÿÿÿÿÿÿÿÿýÿÿÿÿÿÿÿÿüÿÿÿÿÿÿÿÿûÿ

�� øþ��þ� �ü�ü�þ�þ�û��ÿÿÿÿÿ � ÿÿÿÿÿÿÿÿýÿÿÿÿÿÿÿÿüÿÿÿÿÿÿÿÿûÿ

Score = 10/25

0 10 18 25

Score = 20/25

0 10 18 25

Score = 17/25

0 10 18 25

4.4.2.3 Portal deployment benefit

The benefits of a portal deployment are generally substantial [Plumtree, 2001]:

ÿ� More cost-effective, powerful Web infrastructure: Web enables applications and delivers e-business services cost-effectively.

ÿ� Lower network, storage costs: Reduce e-mail distribution of large files by allowing employeesto share documents via a portal, lowering storage and network costs.

ÿ� Lower intranet, extranet administration costs: Allow employees to contribute in a self-serviceway to an organized, secure intranet or extranet, reducing administration costs.

ÿ� Lower training costs: Give people a simple interface to the most useful services from differentapplications, limiting training on complex CRM, ERP and data warehousing clients to thespecialists who use these tools all day.

ÿ� Lower communication costs: Empower employees, partners and customers to get what theyneed from your organization via the portal, rather than by calling someone at a help desk, inhuman resources or at a call center, improving service at lower cost.

ÿ� Improve customer service: Drive more revenues by offering customers a portal into yourbusiness, differentiating your products with a premium electronic resource available 24 hours aday, 7 days a week.

ÿ� Increase productivity: Change the way people work. Improve productivity by giving everyoneyou do business with one place to go to get all of the electronic resources available from yourorganization, minimizing time spent searching or training on complex applications.

ÿ� Improve collaboration, lower travel costs: Use the portal as a collaboration forum foremployees and partners to work together more efficiently without having to meet face-to-faceevery month or every quarter, reducing travel costs.

4.4.2.4 Quantitative studies

To provide some quantitative results about corporate enterprise portal, we resume the results obtained byNucleus Research Inc. [Campbell, 2000], while studying five concrete ROI case studies for differentAmerican companies or public organizations. Within these case studies, Nucleus has looked at the costsand benefits of building a corporate portal using Microsoft® SharePoint™ Portal Server and calculatedthe ROI these five companies or organizations have experienced.


60/100

Aanza Aanza Inc.

One Charing Cross,

Lynnfield, MA 01940.

http://www.aanza.com

Objectives: Building a portal for creating product lifecycle management.

Benefits: Reduction of software development effort,

Shortened time to market,

Increase of reliability,

Reduction of marketing expenses.

Return: Embedded portal application generating positive return,

ROI = 436 %,

Payback period = 3 months.

Children’s Hospital of Boston Children's Hospital Boston300 Longwood AvenueBoston, MA 02115http://web1.tch.harvard.edu/

Objectives: Having a portal as an easier and more scaleable way to post documents.

Benefits: Self serve model to allows the creator to post documents,

Reduction of IT costs,

Increase of accessibility,

Increase of document control.

Return: Portal providing a classic document management solution but leveraging the ease-

of-use and integration of SharePoint with Microsoft Office.

ROI = 409 %,


ARC Advisory Group ARC Advisory Group

3 Allied Drive

Dedham, MA 02026 USA

http://www.arcweb.com/

Objectives: Having a portal to streamline workflows and automate delivery to the web.

Benefits: New document management capabilities,

Complete and flexible workflow for review process,

Shortens review process,

Streamlines publishing to the web.

Return: Formal process to document management.

ROI = 179 %,


Andersan Power Product

http://www.andersonpower.com/

Objectives: Having a portal to get a more efficient way for product teams to develop and share

information.

Benefits: Single place to share documents,

Permanent and rapid document availability,

New broad searching capabilities,

Limited training requirement.

Return: Allow APP to bring new technology into a company traditionally reluctant to

accept change.

ROI = 267 %,



61/100

Rare Medium Rare Medium Group, Inc.

44 W. 18th St., 6th Floor

New York, New York 10011

http://www.raremedium.com

Objectives: Having a portal to standardize on one system and location to share and collaborate

on documents.

Benefits: Single place to share documents,

Permanent and remote availability of any documents,

Broad searching capabilities.

Return: One common platform that enables Rare Medium to centralize documents and

eliminate “islands of information”.

ROI = 187 %,

Payback period = 6.5 months.

Regarding these case studies from Nucleus Research Inc., some common themes may be highlighted inthe use of a corporate enterprise portal, both in terms of functionalities or benefits, as show in table 4.4.

Functionality Benefit areasDocument indexing and categorization,

Portal search and browsing,Tactical collaboration workspaces,

Document-based discussion,Document management.

Improved information organization and access,Improved document collaboration,

Increased tactical collaboration and productivity,Reduced user training and IT support costs.

Table 4.4 - Common themes between the five Nucleus Research case studies on corporateenterprise portals

4.4.3 Specific Ontology enhanced criteria

Regarding the aforementioned portals and knowledge management success functionalities and criteria,some of them may be dramatically enhanced by the use of ontologies. Indeed,ontologies:

ÿ� Improves information and documentation's organization and access,optimising categorization,document management and then, the search and browsing of portal's users.

ÿ� Helps to provide more cost-effective and powerful web portal and knowledge managementinfrastructures. First, it will give to the company's employees the opportunity to bemoreproductive delivering companies and business information more effectively. Second, it willcontribute to amore cost-effective access to e-business servicesfor customers or businesspartners.

ÿ� Contributes to anoptimised accessto documents centralised on the enterprise portal, witheffective query interface and information retrieval capacities. Such a consideration argues infavour of lower network and storage costsfor a company. It will dramatically contribute to thereduction of e-mail distribution of large files between employees, allowing them to sharedocuments via a portal,lowering storage and network costs.

ÿ� Dramatically improves information and document retrieval, delivering to users' requests aunique list of veryefficient answers (documents, information links, etc.). Regarding suchcapacities due to the use of ontologies, employees, partners and customers can get what theyneed directly and efficiently from their organization via the portal, rather than by callingsomeone at a help desk, in human resources or at a call center,improving service at lower cost.

ÿ� Will really improve the productivity of employees in a company, minimizing thetime forsearching information, whenever they want and wherever they are.


62/100

4.5 E-Commerce

4.5.1 Reference evaluation methodologyNot known yet.

4.5.2 Reference evaluation criteriaNot known yet.

4.5.3 Specific Ontology enhanced criteriaNot known yet.


4.6.1 Reference evaluation methodologyNot known yet.

4.6.2 Reference evaluation criteriaNot known yet.

4.6.3 Specific Ontology enhanced criteriaNot known yet.

4.7 References

[Salton 1983], G. Salton, M.J. McGill "Introduction to modern information retrieval ", - McGraw-Hill publishing company 1983.

Especially the chapter on "Retrieval Evaluation".

[Ricardo et al., 1999], Ricardo Baeza-Yates, Berthier Ribeiro-Neto, "Modern Information Retrieval ", ACM press, Addison

Wesley, 1999. Especially the chapter on "Retrieval Evaluation".

[SemWeb-ercim] web sitehttp://www.ercim.org/EU-NSF/semweb.htmlespecially the slides onOntology-enhanced retrieval(and

ontology-enhanced applications) Deborah L. McGuinness, Stanford.

[FindUR] Ontological Issues for Knowledge-Enhanced Search, deborah L. McGuinnesshttp://www.research.att.com/~dlm/findur

[Chandrasekaran et al.; 99] Chandrasekaran, B.; Johnson, T. R.; Benjamins, V. R.Ontologies: what are they? why do we need

them?. IEEE Intelligent Systems and Their Applications. 14(1). Special Issue on Ontologies. 1999. Pp. 20-26.

[Gruber, 93] Gruber, T. R.A translation approach to portable ontology specification. Knowledge Acquisition. #5: 199-220. 1993.

[Guarino et al.; 95] Guarino, N.; Giareta P.Ontologies and knowledge bases: Towards a terminological clarification. In N. J. I.

Mars, editor, Towards Very Large Knowledge Bases: Knowledge Building & Knowledge Sharing. IOS Press. Amsterdam

(Netherlands). 1995. Pp. 25-32.

[Borst, 97] Borst, W. N.Construction of Engineering Ontologies. University of Tweenty. Enschede, NL- Centre for Telematica and

Information Technology. 1997.

[Studer et al.; 98] Studer, R.; Benjamins, V.R.; Fensel, D.Knowledge Engineering: Principles and Methods. Data & Knowledge

Engineering. 25: 161-197. 1998.

[Gómez-Pérez, 2002] Gómez-Pérez A., "Evaluation of Ontologies", To appear in International Journal of Human ComputerStudies, 2002.

[Gómez-Pérez, 1996] Gómez-Pérez A., "A Framework to Verify Knowledge Sharing Technology", Expert Systems withApplication, Vol.11 (4), pp519-529, 1996.

[Gómez-Pérez, 1994] Gómez-Pérez A., "Some ideas and Examples to Evaluate Ontologies", Technical Proceedings of the 11thConference on Artificial Intelligence for Applications. CAIA94, 1994.

[Trec] http://trec.nist.gov


63/100

5 Successful Scenarios and Guidelines

5.1 IntroductionSuccessful scenarios and Guidelines are documented strategies and process employed by top-of-classApplications. One given application can not claimed to be top-of-class in every area - such applicationdoes not exist - so in D2.2 and series, we are capturing the best practices of the top best applications in agiven class.

5.2 Corporate Intranet and Knowledge ManagementIn [Dieng et al, 2001], the following guidelines are presented. The corporate memory can be materializedby “a corporate semantic Web”, constituted of ontologies, documents (possibly XML documents) andsemantic annotations of these documents, these annotations using the conceptual vocabulary defined inthe ontologies. Such a memory will be a hybrid memory both document-based and knowledge-based.[Dieng et al, 2001] proposes thefollowing methodology for construction of such a memory:

ÿ� For each type of document to integrate in the memory, choose a DTD indicating the model to berespected. For example, there may be a DTD for defining structured business dossiers, another onefor experience forms, another one for project forms, etc.

ÿ� The authors of documents can then create their XML documents, respecting the DTD adopted forthe considered type of document.

ÿ� Several XSLT style sheets can be defined so as to have different presentations of documents,according to the considered users.

ÿ� XLink can be exploited to define hypertext links from the external of a document, without touchingthis document. It will enable to have several annotations according to several viewpoints on thesame document. These hypertextual annotations can thus depend on the users and on the context. Itwill enable to take into account differences among professions or activities of users, inside anenterprise.

ÿ� The members of the enterprise can agree on an ontology and represent it in the appropriate KR. Thisontology will be more or less detailed according to its objectives: either it aims only to enableannotations of documents or it aims to be a component of the corporate memory and to be consultedas such. According to the available information sources (persons, documents, structured or semi-structured databases) and according to the degree of detail, width and depth of this ontology, severalmethods and techniques can be used for construction of this ontology:

ÿ� Construction of an ontology and of an enterprise model: One can use the method proposed in theCoMMA project and then organize the ontology in several levels: (1) a generic level inspired fromhigh-level ontologies, (2) a central level with a part linked to corporate memory and the other one onthe domain subject, (3) a layer specific to the intended use scenarios and to the enterprise.

ÿ� Construction from several experts: If a rather fine modeling of the different experts is needed,knowledge acquisition methods from several experts can be used. The ontology can be organizedthrough several viewpoints. In case of expertise conflicts, one can adopt different conflictmanagement strategies.

ÿ� Construction from documents: In the case when knowledge sources are textual, one can use themethod of construction of an ontology from texts (Aussenac-Gilles et al, 2000) integrating NaturalLanguage Processing (NLP) Tools. In the case of a project memory and, especially for building thememory of problems encountered in a project, one can rely on the method proposed by SAMOVARfor building the ontologies, by using some documents or textual data in existing databases.

ÿ� Once the ontology established, one can use the conceptual vocabulary of the ontology so as to writethe semantic metadata about the documents of the memory. Let us notice that if these semanticannotations are described in files separate from the documents to be annotated, it enables to keep the


64/100

corporate legacy documents in their present form (i.e. without necessarily transforming them intoXML documents), and to retrieve them via their semantic metadata stored separately. In case ofneed, one can use NLP tools so as to automate (at least partially) the annotation process.

ÿ� In the enterprise, it may be needed to distinguish several communities corresponding, for instance,to different professions inside the organization (e.g. administrative services, financial services,engineering centers, designers…). In this case, one can possibly build different ontologies, one foreach community; each document can then have different annotation contexts, corresponding to thedifferent intended communities. Another possibility would be to build a multi-terminologiesontology organized in several viewpoints.

ÿ� For exploitation of the memory, the users will need to look for information about the documents.The use of a reasoning engine will enable them to perform document retrieval guided by semanticcriteria, by using the semantic metadata.

ÿ� Another possible use of the memory consists of advising in a proactive way the users about thepresence of new documents that may interest them (for example documents upon which annotationsindicate they may be interesting for such or such user community or that they are aimed at usersfrom this or this profession or having this or this interest center).

ÿ� A feedback from the users and statistics about the use of the memory will enable to help toevaluation of the memory and to plan its evolution.

ÿ� According to the organizational choice, the evolution of the memory components (ontologies,documents or annotations) will be centralized or distributed.

Remark: One can consider that the base of annotations constitutes a network enabling to browse thedocument space if the documents are the main component of the memory. But the memory can also allowaccess to information non via documents but via persons. In this case, the memory components will bethe experts, the ontologies and annotations about these experts, using the conceptual vocabulary of theontologies. This approach aims at automating the contact with experts that can support a person or agroup and it favors the emergence of interest communities.

5.2.1 Guidelines[Rose Dieng et al 2001] in a conclusion confirms the multiple research fields relevant for building acorporate KM - which definitively requires a multidisciplinary approach. The choice between thedifferent construction techniques can be based on several questions that an enterprise should answerbefore building a corporate KM:

1. Needs detection:

ÿ� Design a knowledge-management system based on user needs, requirements and usage habits.

ÿ� Who are the potential users of the KM and what are the users’ profiles?

ÿ� What is the intended use of the KM after its construction?

ÿ� A mean of communication between distant groups?

ÿ� A mean of communication between an enterprise and privileged partners?

ÿ� A mean to enhance learning of new enterprise members?

ÿ� When will the KM be used: in short-term, in mid-term, or in long-term?

2. Construction:

ÿ� Choose a system that is consistent with your organization's culture and user patterns.

ÿ� Work with existing systems. Do not duplicate existing resources.


65/100

ÿ� What are the knowledge sources available in the firm: paper-based, semi-structured or structureddocuments, human specialists, and databases?

ÿ� Can the quality, volume, availability of the knowledge sources be assessed?

ÿ� What is the knowledge map of the enterprise departments involved in the knowledge managementoperation?

ÿ� What kind of knowledge must contribute to the construction of the Corporate KM?

ÿ� Knowledge already described in documents such as reports or synthesis document on a project?

ÿ� Elements of experience and professional knowledge not already described in documents?

ÿ� Is it necessary to model knowledge of some enterprise members or is an intelligent documentarysystemsufficient?

ÿ� What is the preferred materialization, according to the computer environment of both future usersand developers and according to the financial, human and technical means available for theCorporate KM construction and maintenance?

ÿ� Organize information simply: "Complexity discourages usage"

ÿ� Design a system that adds value to information.

ÿ� Set a realistic budget to develop and maintain the system. Then adhere to the budget.

3. Diffusion:

ÿ� Seek top management support for the system and ensure a high-level champion actively promotesand encourages use of the system.

ÿ� What is the preferred scenario of interaction between the future users and the Corporate KM?

ÿ� What interface will be the most adapted to the users’ activity environment?

ÿ� What will be the privileged diffusion means (Internet, Intranet,...), according to the computerenvironment of both future users and developers ?

ÿ� Encourage and train people to use the system.

4. Evaluation:

ÿ� What will be the evaluation criteria?

ÿ� When, how and by whom will such an evaluation be carried out?

5. Evolution:

ÿ� How will the evaluation results be taken into account?

ÿ� When, how and by whom will the Corporate KM be maintained, verified and incremented?

ÿ� How will obsolete or inconsistent knowledge be detected and removed (or conceptualized)?

ÿ� Will a department centralize the evolution of the Corporate KM or will it be distributed amongseveral members of the organization? In the second case who will coordinate the effort?

ÿ� Frequently updated information in the system. Designate an individual to manage and maintain thesystem.

5.2.2 Keys to the success in building an enterprise portal

Measuring and maximizing return on investment (ROI) appears as a critical issue and decisional factor fora company that is thinking of building an enterprise portal, often based on knowledge management,


66/100

information retrieval and/or ontology technologies. According to Campbell, the keys to success inbuilding an enterprise portal (mainly in terms of ROI maximization) can be summarized in the followingway.

1- First, it is very important for the company not to think about an ideal portal, but to think about a portalaiming at answering some well-identified needs or business objectives.

2- Second, it is important to previously look and evaluate the expected breadth of the portal inconsistency with the business objectives and its repeatability.

3- Third, the different technology options that may be integrated within the portal must be evaluated interms of:

ÿ� Support for business objective(s),ÿ� Rapidity for development and deployment,ÿ� Level of training cost for employees of the company.

4- Fourth, when the portal is deployed, it is from crucial importance for the company to communicate onthe portal’s objectives, and highlighting the fact that users (employees for B2E, customers for B2C orbusiness partners for B2B) must now take benefits of the use of the portal as a new efficient informativeand collaborative tool, which will also help them to earn some time and being more productive.

5- Fifth and finally, it is crucial to evaluate the real use of the portal and estimate its benefits. Then, theobtained results must be analyzed so that conclusions can draw the success of the portal or the need ofnew evolutions to really suit to the initial business objectives and needs for users.

Moreover, it is generally difficult tocalculate ROI. However, some few important points may be kept inmind when embarking on a ROI calculation:

ÿ� Use a 3-year horizon and calculate ROI using the average savings over the 3-year period dividedby the initial cost.

ÿ� Payback period is an indicator of risk.ÿ� Don't get misled and include costs that are not associated with the project.ÿ� There are only six categories of cost:

ÿ� Software, ÿ� Consulting,ÿ� Hardware, ÿ� Training,ÿ� Personnel, ÿ� Other.

ÿ� Keep in mind that some costs (and savings) are one-time while others are recurring.ÿ� Savings can be direct and indirect.ÿ� Correct productivity gains for inefficient transfer of time.ÿ� Choose a consistent methodology and apply it to every technology decision.

Using these aforementionedfinancial metrics, a company gets the right decision inputs for maximizingits return. Such inputs can also be applied in a more general way when companies have to estimate theadded value of the deployment or use of new technology.


67/100

5.3 E-Commerce

5.3.1 Guidelines for E-Business: B2C and B2B

Companies [Best Practices LLC] that fully integrate the Internet into their overall strategy and existingbusiness operations are experiencing dramatic productivity gains and marketplace advantages. Thosemanagers and companies makingdramatic progress in e-marketing, e-sales and e-services-- observe itis only a matter of time before the Internet is completely integrated into every facet of their overallstrategy and business operations. Then, these e-commerce leaders reflect, today's "e-business" will betomorrow's "business as usual." This study is anchored on four critical perspectives that serve as thecornerstones for fieldwork, analysis and organization of all insights. These critical perspectives includethe following:

ÿ� The Internet has emerged as an essential tool to enable and supercharge productivity.

ÿ� The Internet can and should be used to enhance existing successful business models.

ÿ� Internet managers must translate e-business tools, techniques and initiatives into real stakeholdervalue that can be calculated in traditional terms of business economics and productivity.

ÿ� Broad-based Internet strategies and initiatives can be analyzed and understood in terms of specificpractices that are repeatable and manageable throughout multiple enterprises.

5.3.1.1 Global key Findings

Five Key Findings that have been reported by [Best Practices LLC] follow:

ÿ� 1- Evaluateyour business model and apply Internet tools to the highest value points.

ÿ� 2- Integrate e-tools with traditional customer service excellence principles to drive long-termloyalty, faster responses, greater ease of use, increased functionality and cost reduction.

ÿ� 3- Enhancesales productivity by integrating Internet tools with existing sales channels.

ÿ� 4- Employ e-marketing to gather information and to segment and target customers with mass-customized communication.

ÿ� 5- Measuree-business success based on deployment of people, technology and strategy to accelerateimprovement cycles and increase profit margins.

According to [Dubosson-Torbay et Al., 2001], “Business model“ is one of the latest buzzword in theInternet and electronic business world. Business models may become a major stake in e-commerce andbusiness. A business model is nothing else than a formalized and understandable representation of acompany and its partners (B2B or B2C) for creating, marketing and delivering value and relationshipcapital to one or several segments of customers, in order to generate profitable and sustainable revenuestream. In that way, it clearly appears that a company who wants to estimate or precisely evaluate the(potential) added value of the introduction of ontology or knowledge-based technologies in its e-commerce application must try to directly integrate these technologies as input in its whole businessmodel, in consistency with its organizational goals and business objectives. Thus, the added value of theuse of such knowledge technologies may be estimated by a company since a comparison betweengenerated outputs of the business models integrating or not ontologies becomes possible.

At the present time, most of the studies dealing with eBusiness model design, classification andmeasurements, are trying to define some critical success factors, in order to find out and compare theperformance indicators used by eBusiness companies which are competing with similar business models.Our interest on that subject within our study is larger than that, trying to compare and estimate thegenerated outputs of a company whether it uses ontology-based technology or not.


68/100

If considering the eBusiness model decomposition provided in [Dubosson-Torbay et Al., 2001], the useof ontology-based technologies may be introduced at different level as shown on figure 5.1.

Figure 5.1 - eBusiness model decomposition

Red circles focus on the potential “location” of knowledge-based technologies, such as ontology, on theeBusiness model of a company, to evaluate the added value of its introduction in terms of financialrevenues or gains of productivity, at the B2E, B2B or/and B2C levels.

5.3.1.2 Ontology based B2C

Whilst many private initiatives have used ontologies as part of B2C applications there has, to date, beenlittle effort to formalize best practice within the B2C electronic commerce applications. However, theOntology Based technology should bring added value to the classical Value Chain:

Figure 5.2 - Classical chain value for ontology-based technologies.

CustomerRelationship Value for

ProductInnovation

Resources for InfrastructureManagement

Information

Feel & serve

Trust & loyalty

Information

Value proposition

Capabilities

Resources/ assets

Activities/processes

Partner network

Revenue Value added Costs+

Price Profit


69/100

5.3.1.3 Ontology based B2B

[CEN/ISSS] Classification schemes have a long history within electronic commerce. The most influentialof these schemes have been developed under the auspices of United Nations’ Electronic Data Interchangefor Administration, Commerce and Transport (UN/EDIFACT) initiative by the United Nations Centre forTrade Facilitation and Electronic Business (UN/CEFACT).Other long-standing classification schemes that are relevant to E-Commerce include:

ÿ� The Universal Standard Product and Services Classification (UNSPSC) developed by theElectronic Commerce Code Management Association (ECCMA)

ÿ� ISO 4217: Codes for the representation of currencies and funds developed by the InternationalOrganization for Standardization

ÿ� The International Standard Industrial Classification of All Economic Activities (ISIC) developedby the UN Statistics Office.

At present, however, established standardization bodies have not chosen to develop formalized ontologiesas basis for classification schemes.Best practices relating to the use of classification schemes within B2B electronic commerce are currentlybeing studied and standardized under the auspices of UN/CEFACT and OASIS as part of the ebXMLinitiative and the follow-on ebTWG and UBL work. Details of this on-going work can be found athttp://www.ebtwg.organdhttp://www.oasis-open.org/committees/ubl/.A wide range of industry specific ontologies are being prepared, among the more advanced of which arethe ones specific to the electronic component supply industry (RosettaNet) and the automotive industry(AIAG). Details of the current practices of both these groups can be found at their websites,http://www.rosettanet.organdhttp://www.aiag.org.[Feindt et al. 2001] Once an SME has drawn up a strategy for supporting B2B e-commerce interactions, itcan determine the factors critical to success in meeting this strategy. Critical success factors are defined as‘the limited number of areas in which results, if they are satisfactory, will ensure successful competitiveperformance for the organization.’ [Chappell, 2000] Based on previous research and the 43 SMEs of theKITS sample we have identified the followingsix critical success factors:

5.3.1.4 Critical success factors in B2B

# Guideline Item Issues

1 Vision

Understanding that an SME’s competitiveness in its industryvalue chain depends on external efficiencies as well as internalones – that is, efficient interactions with customers andsuppliers

2 CultureEncouraging open access to information and collaborativeprocesses across internal and external boundaries

3 TrustEncouraging value chain partners to feel as high a level oftrust in the SME as they would in their own organizations

4 Value

Ensuring that the B2B e-commerce interaction between valueactivities (for example, a customer’s procurement activity anda supplier’s sales and marketing activity) is innovative andadds value over traditional forms of interaction (such as paper,phone or fax)

5 Control

Establishing the appropriate level of control over the valuechain interaction, depending on factors such as businessenvironment, technological capability of the partners, criticalnature of the value chain interaction

6 IntegrationProviding internal links between the organization's IT systemsand between its IT systems and those of its partners


70/100

5.3.1.5 Ontologies and SMEs

Small and Medium-sized Enterprises (SMEs) do not have sufficient internal skills or resources to developand maintain complex ontologies. Typically their staff will lack language skills, often only being able tooperate efficiently in a single language. Often the size of an SME’s market is limited by this lack ofknowledge, and they find moving into new areas within the European single market difficult and oftendaunting.

SMEs look to professional organizations, such as trade associations, to provide shared resources that canbe used to standardize tasks carried out in many different companies. They rely on such bodies to developthe appropriate consensus as to the best way to categorize and describe products, services and theirproviders. SMEs require that products and services be developed in conformance with agreedcategorizations that they can install as part of their business systems with the minimum of fuss, and havetheir staff use with the minimum of training.

Trade associations typically deal with a single industry in a single country. As the European SingleMarket develops there will be a need to develop pan-European industry associations that can preparemultilingual ontologies that allow businesses to work in any European country. There will also be a needfor the integration of ontologies developed for specific industries into generalized ontologies for productand service description. This latter is necessary for a number of reasons, including:

ÿ� The need to reduce the amount of duplicated effortÿ� The need to avoid multiple sets of terms for describing the same thingÿ� The need to reduce costs by reusing software components across multiple industries.

Many industries share the same processes. Many industries share the same components. Many industriessupply the same customers. If each industry introduces its own techniques for describing processes,components and customers not only will it be impossible to develop reusable software components, butalso it will be impossible to harmonize work within companies serving multiple industries.

Software suppliers would love to be able to dominate a particular market, providing products and servicesthat are a necessary part of a particular industrial process. They will naturally try to suggest that theontology they have developed as a core component of their system is the best available, and should beadopted by all players. They will typically not be interested in how their ontology can be integrated withthose used in other industries/processes. SMEs cannot easily apply pressure to software suppliers toensure that the products they are purchasing conform to industry norms. They must rely on someone whocan take up the cause of many different SMEs to take on the task of ensuring that the ontologies usedwithin products do set suitable norms that conform to working practice within each of the communities inwhich they trade. To do this they must first make their representatives in the trade associations taking upthis task aware of their needs in terms of coverage of relevant ontologies, particularly with respect tolinguistic requirements.

User associations that are to help develop multi-industry multilingual ontologies will need to employlinguists that are familiar with the terms used in each of the industries being covered by the ontologiesthey develop. These linguists will need to be supported by panels of users who are able to provideprofessional evaluation of the results, and answer any questions that might arise during the production ofthe ontologies. Checking the relationships between different terms will require expertise in the way inwhich terms are used differently within different languages. A mechanism needs to be put into place toallow errors identified in ontologies to be reported in such a way as to correctly capture the correctrelationships.


71/100

Relationships between the terms used to describe products, services and processes within differentlanguages are rarely exact. Sometimes there will be one-to-many relationships between terms. Sometimesa term will mean more in one language than its equivalent in another language. Sometimes there will beno equivalent of a term within a particular language. Ontologies need allow for the recording of therelationships between terms, and for the identification of many-to-many relationships betweenpolysemous terms. These relationships should be defined using terms that are understandable to SMEsrather than in terms only understandable to trained linguists so that SMEs can efficiently reviewontologies.


72/100

5.4 Information Retrieval

The last few years have seen a dramatic increase in the generation of information as a result of highthroughput technologies in certain leading edge fields (e.g. in the genomics field with the reading of thehuman genome) but also as a consequence of the ease with which information can be published.

A result of this explosion is that the extraction of knowledge from information, and not the availability ofinformation itself, has become a critical factor for business. From a business perspective therefore,effective information retrieval has become an essential element of the competitive capability ofcompanies within knowledge-intensive industries. Furthermore, the central role that information plays indecision-making creates the need for high quality of the retrieved information.

Most technologies that have been used to date have paid lip service to quality issues partly because otherperformance metrics (such as response times) were of higher priority. This however is no longer the caseand instead quality issues (such as relevance, accuracy, completeness, conciseness) are now important.

Ontology-based approaches promise to increase the quality of responses since they aim to capture withincomputer systems some part of the semantics of concepts allowing for better information retrieval.However, while the opportunities for value creation exist, the path is anything but straight.

Amongst the opportunities and risks that exist, the main ones are the following:

ÿ� Opportunities:o Data is being created at phenomenal rates. It is felt that information (knowledge) is hidden

within this data and companies are under continuing and increasing pressure to extract valuefrom it,

o Better information retrieval means better decision-making,o Identification of business opportunities (matching buyers with sellers),o Enhanced problem solving (and scientific discovery) capability,o Market is still fragmented both on the supply and the demand side. This creates

opportunities for suppliers that will address some of the consequences of fragmentation suchas system interoperability, supply chain integration and others.

ÿ� Risks:o The use of proprietary ontologies for all critical applications creates the need for high

quality ontology alignment. Failure to achieve this will put the vision of systeminteroperability at risk,

o Ontologies must be flexible, open ended and capable of dealing with multiple definitions.Failure to achieve this will jeopardize the acceptance of ontology-based systems.

o In many cases the business models that are being adopted are not clear and result in failuresthat can be blamed on the technology. Ontology based applications, like any technicalsolution, must be well integrated within the business processes of an industry and must beshown to add value to the existing value chain or creating a new, viable value chain.

5.4.1 Guidelines

The aim of this section is to build on the framework discussed in previous sections and provides apractical checklist of issues to examine when considering the implementation of an ontology-basedapplication.

While the nature of the application will influence in large measure the development and especially thedeployment paths followed, the following list provides generic guidelines that hold in most cases.


73/100

The list focuses not so much on the technicalities of the implementation but more on issues that relate tothe business environment that affect the deployment, integration and acceptance of the ontology-basedapplication witha view to maximize the business value extractedfrom it.

Issues such as maintenance, IP, difficulty of use, knowledge sharing, and data noise often impinge on thesuccessful integration of a new technology within an organization and yet are given only lip service. The‘absorptive capacity’ of an organization for a new technology, i.e. its ability to integrate within existingpractices and utilize effectively new systems, is becoming critical to its competitive advantage especiallyin the current environment of rapid and widespread technological change.

Table 5.1 below lists guidelines for the development and deployment of ontology-based IR applicationsand should be used at the design stage of such a system.

# Guideline Item Issues

1Decide on the nature of the

application

Will the ontology be used forÿ� Descriptions?ÿ� Integration of heterogeneous info systems?ÿ� Better information retrieval?ÿ� Support of problem solving?

Based on the intended use, a different set of evaluation criteriashould apply.

2Identify users and involve them

from start

Because ontologies are about definitions and “understanding”between communicating parties, where humans are to beinvolved they constitute a fundamental issue. To improvetransparencyand hence acceptance of the system, its intendedusers must be consulted from the very beginning.

3Look at company

information/knowledge ‘silos’

Often groups within large organizations exist in “silos” wherecommunication between them is non-existent or not veryeffective. If this is the case, are there duplicate databases /ontologies and how can the ontology application help bridgethe gaps between the silos?

4Decide on proprietary and non-

proprietary concepts(parts of ontology)

If the intended application is of strategic importance, are theresome parts of the ontology that must be kept companyconfidential? If this is the case, is a partitioning between theproprietary and public parts of the ontology possible? Can thenon-proprietary part be found in the public domain and whatare the implications in terms of integration with theproprietary part?

5 Examine your legacy systems

If the ontology application is not custom built check thatintegration with legacy systems is possible. This should gobeyond the communications layer and consider the logicallayers of the legacy and ontology systems.

6 Decide on Reasoning Processes

What kinds of use will the ontology be put to? If it to be usedfor reasoning and problem solving, are there any reasoningengines available or would it be necessary to develop one in-house? What is the state of the art and what the level ofmaturity of any existing technologies?

7Define evaluation criteria and

metrics

Although success will be most probably proved in practice, itis important to be able to quantify the advantages of the newsystem and hence its ROI. This should include thetotal effortinvested and required for developing and maintaining the newsolution.

8 Decide whether to cover all Information retrieval is not always performed with the same


74/100

search modes goals in mind and the range of expected / useful answersmight vary considerable for the same query. Each moderequires different indexing and scoring functions that shouldbe explicitly encoded in the system from the beginning.

9Consider how to deal with data

noise

There are many applications that must assume noisy data (e.g.fault repair records, call center records etc.) with non-standardgrammar and syntax (e.g. short hand). This noise couldseverely affect system performance and therefore should alsobe accounted for from the beginning of the process.

10 Research Existing Ontologies

Does an existing ontology exist that covers your applicationneeds? This possibility should be considered seriously if theapplication will need to make use of industry wideinformation norms and if the information is not considered ofstrategic value to the company.

11Look at Ontology Maintenance

tools and procedures

If a specific ontology environment (that accompanies theontology) is to be used examine the capabilities of thesupporting tools, the assumptions of the underlying languageand its expressive power. How easy is it to change conceptdefinitions? Does the environment check for inconsistencies?Can it cope with multiple versions? Etc.

12If you are to use externalontologies, look at whomaintains / backs them

Is the external ontology you have decided to use approved /backed by any industry wide body? Is there a maintenanceprocedure? Is there a working group that maintains theontologies and allows users to voice their needs? Are otherindustry participants using this ontology?

13Consider integration with yoursupplier / customer applications

Is there any benefit in integrating the ontology applicationwith that of your customers or suppliers? If so, do they haveany equivalent applications in place? Is integration with yourontology possible and at what level?

Table 5.1 -Checklist for the implementation of an ontology application

Several experiments have been carried out in the last 15 years investigating the use of various resourcesand techniques (e.g., thesauri, synonyms, word sense disambiguation, etc.) to help refine or enhancequeries. However, the conclusions drawn on the basis of these experiments vary widely. Results of somestudies have led to the conclusion that semantic information serves no purpose and even degrades results,while others have concluded that the use of semantic information drawn from external resourcessignificantly increases the performance of retrieval software. At this point, several question arise:

ÿ� Why do these conclusions vary so widely?ÿ� Is the divergence a result of differences in methodology?ÿ� Is the divergence a result of a difference in resources? What are the most suitable resources? Do

results using manually constructed resources differ in significant ways from results usingautomatically extracted information?

ÿ� What is the contribution of specialized resources?ÿ� Are present frameworks for evaluation (e.g., TREC) appropriate for evaluation of results?

These questions are fundamental not only to research in document retrieval, but also for informationsearching, question answering, filtering, etc. Their importance is even more acute for multilingualapplications, where, for instance, the question of whether to disambiguate before translating isfundamental.


75/100

Several experiments are trying to bring together researchers in the domain of document retrieval, and inparticular, researchers on both sides of the question of the utility of enhancing queries with semanticinformation gleaned from languages resources and processes.

In view of the relevance of Word Sense Disambiguation (WSD) to information retrieval, a part of theworkshop will be dedicated to a panel on Senseval and WSD evaluation(http://www.itri.bton.ac.uk/events/senseval/).

The purpose of the panel is to open the discussion on the objectives and principles of Senseval-3, the nextevaluation exercise for WSD systems.

The panel will summarize the methodology and results of Senseval-2 and consider the current state ofWSD. Panellists will speak about the following topics:

ÿ� Is the Senseval-2 method of evaluation adequate?ÿ� Are we building the right kind of lexical resources for real sense disambiguation systems?ÿ� Should we (and is it possible to) start an application-specific evaluation track?


76/100


5.5.1 GuidelinesLessons Learntfrom the SEAL Representative application [Stajanovic et al. 2001]

The conceptual backbone of the SEAL approach is an ontology. For the AIFB portal, they had to modelconcepts relevant in this setting. As SEAL has been maturing, a methodology was developed for settingup ontology-based knowledge systems [Staab S. et al. 2001]. The approach (cf. figure 5.3) is mainlybased on [Schreiber B. et al. 1999] and [Gomèz-Pérèz A. 1996] but focuses on the application-drivendevelopment of ontologies. Here is described some experiences made during the ontology developmentfor the SEAL portal.

Figure 5.3 - Ontology Development : global picture

Kickoff phase for ontology developmentSetting up requirements for the SEAL ontology the SEAL team had to deal mainly with modeling theresearch and teaching topics addressed by different groups of the AIFB institute and personal informationabout members of the AIFB institute. The SEAL team took themselves as an “input source” and collecteda large set of lexical entries for research topics, teaching related topics and personal information, whichrepresent the lexicon component of the ontology. By the sheer nature of these lexical entries, the ontologydevelopers were not able to come up with all relevant lexical entries by themselves. Rather it wasnecessary to go through several steps with domain experts (viz. institute colleagues) in the refinementphase.

Refinement phase.We started to develop a baseline taxonomy that contained a hierarchy of research topics identified duringthe kickoff phase. An important result for the team was to recognize that categorization was not based onan isA-taxonomy, but on a much weaker HASSUBTOPIC relationship.E.g. “Knowledge Discovery inData Bases” is a subtopic of “Knowledge Management”, which means that it covers some aspects of“Knowledge Management” —but it does not reflect inheritance provided by an isA-taxonomy.

It then took the team three steps to model the currently active ontology. In the first step, all memberscollected lexical entries from the institute. Though the team had already given the possibility to provide arough categorization, the categories modeled by non-knowledge engineers were not oriented towards amodel of the world, but rather towards the way people worked in their daily routine. Thus, theircategorization reflected a particular rather than a shared view onto the domain. A lesson learned from thiswas that people need an idea about the nature of ontologies to make sound modeling suggestions. It wasvery helpful to show existing prototypes of ontology-based systems to the domain experts.

In the second step, the team worked towards a common understanding of the categorization and thederivation of implicit knowledge, such as “someone who works in logic also works in theoretical

ONTOLOGY

Ontologykickoff

EvaluationRefinement Maintenance


77/100

computer science” and inverse of relations,e.g.“an author has a publication” is inverse to “a publicationis written by an author”.

In the third step, the team mapped the gathered lexical entries to concepts and relations and organizedthem at a middle level. Naturally, this level involved the introduction of more generic concepts thatpeople would usually not use when characterizing their work (such as “optimization”), but it alsoincluded “politically desired concepts”, because one own’s ontology exhibits one’s view onto the world.Thus, the ontology may become a political issue!

Modeling during early stages of the refinement phase was done with pen and paper, but soon the teamtook advantage of the ontology environment OntoEdit [see Onto-To-Knowledge] that supports graphicalontology engineering at an epistemological level as well as formalization of the ontology. Like in otherontology engineering projects, the formalization of the ontology is a non-trivial process.

Evaluation phase.After all the team found that participation by users in the construction of the ontology was very good andmet the previously defined requirements, as people were very interested to see their work adequatelyrepresented. Some people even took the time to learn about the Ontology editing tool "OntoEdit".However, the practical problem the team had was that our environment does not yet support an ontologymanagement module for cooperative ontology engineering.We embedded the ontology into our institute portal. It contains around 170 concepts (including 110research topics) and 75 relations. This version is still running, but we expect maintenance to be a relevanttopic soon. Therefore the team is collecting feedback from our users —basically colleagues and studentsfrom the institute.

5.6 Other approaches to be used to develop domain ontologies

Other approaches that could be used to develop domain ontologies exist :

Uschold and King's approach [Uschold et al, 1995]: They proposes: i) to identify the purpose of theontology, ii) to build it, iii) to evaluate it, and iv) to document it. During the building phase, they proposecapturing knowledge, coding it and integrating other ontologies inside the current one.

Grüninger and Fox's approach [Grüninger et al, 1995]: The authors propose a formalized method forbuilding ontologies. First, they propose to identify intuitively the main scenarios (possible applications inwhich the ontology will be used). Later, a set of natural language questions, called competency questions,are used to determine the scope of the ontology, that is, the questions that could be answered using theontology. These questions are used to extract the main concepts, their properties, relations and axioms,which are formally defined in Prolog.

KACTUS approach [Bernaras et al, 1996]: The ontology is built on the basis of an applicationknowledge base (KB), by means of a process of abstraction (that is, following a bottom-up strategy). Themore applications are built, the more general the ontology becomes; hence, the further the ontologymoves away from a KB. In other words, they propose to start building a KB for a specific application.Later, when a new knowledge base in a similar domain is needed, they propose to generalize the first KBinto an ontology and adapt it for both applications. Applying this method recursively, the ontology wouldrepresent the consensual knowledge needed in all the applications.

Methontology [Fernández-López et al,1999]: It is a methodology for building ontologies either fromscratch, reusing other ontologies as they are, or by a process of reengineering them. The Methontologyframework enables the construction of ontologies at the knowledge level. It includes: identification of theontology development process, a proposed life cycle and the methodology itself. The ontologydevelopment process identifies which tasks should be performed when building ontologies (planning,


78/100

control, quality assurance, specification, knowledge acquisition, conceptualization, integration,formalization, implementation, evaluation, maintenance, documentation and configuration management).The life cycle, which is based on evolving prototypes, identifies the stages through which the ontologypasses during its lifetime, as well as the interdependencies with the life cycle of other ontologies. Finally,the methodology itself specifies the steps to be taken to perform each activity, the techniques used, theproducts to be output and how they are to be evaluated. The main phase in the ontology developmentprocess using the Methontology approach is the conceptualization phase. During both specification andconceptualization, a process of integration was completed using in-house and external ontologies. Thisframework is partially supported by the software platform WebODE [Arpírez et al.; 2001]. Methontologyhas been recommended by the Foundation for Intelligent Physical Agents (FIPA) for ontologyconstruction (http://www.fipa.org/). Moreover, Methontology is the methodology that has mostcompliance with the IEEE standard for software development [IEEE, 1996].

Reader is invited to consult [Fernández-López, 1999] to get study on ontology developmentmethodologies.


79/100

5.7 References in Guidelines

Knowledge Management[Dieng et al., 1999], R. Dieng, O. Corby, A. Giboin, and M. Ribiere.Methods and tools for corporate knowledge management.


[Abecker et al., 2002] A. Abecker, A. Bernardi, S. Dioudis, L.Van Elst, R. Herterich, C. Houy, M. Legal, G. Mentzas, S. Müller, G.Papavassiliou.Enabling Workflow-Embedded On Access With The Decor Toolkit.To appear in R. Dieng-Kuntz and N. Mattaeds, Knowledge Management and Organizational Memories, Kluwer, 2002.

[Dieng et al., 2001] R. Dieng, O. Corby, F. Gandon, A. Giboin, J. Golebiowska, N. Matta, M. Ribière,Méthodes et outils pour lagestion des connaissances : une approche pluridisciplinaire pour le "knowledge management", 2nd édition, Eds. DUNOD,INFORMATIQUES, Série Stratégies etsystèmes d'information, 2001.

[Van Elst, 2002] L. Van Elst, A. Abecker,Domain Ontology Agents In Distributed Organizational Memories. To appear in R.Dieng-Kuntz and N. Matta eds, Knowledge Management and Organisational Memories, Kluwer, 2002.

[Gandon, 2001] F. Gandon,Engineering an Ontology for a Multi-Agents Corporate Memory System, Proc. of the EigthInternational Symposium on the Management of Industrial and Corporate Knowledge, Technological University of Compiègne,France, 22-24 October 2001.

[Golebiowska et al, 2001] J. Golebiowska, R. Dieng-Kuntz, O. Corby, D. Mousseau,Samovar : Using Ontologies And Text-Mining For Building An Automobile Project Memory , Proc. of K-CAP, Victoria, October 2001.

[Kalfoglou, 2002] Y. Kalfoglou,Maintaining Ontologies With Organisational Memories And Supporting OrganisationalMemories With Ontologies. To appear in R. Dieng-Kuntz and N. Matta eds, Knowledge Management and OrganisationalMemories, Kluwer, 2002.

[Staab, 2002] S. Staab, H.P. Schnurr,Knowledge And Business Processes: Approaching An Integration. To appear in R. Dieng-Kuntz and N. Matta eds, Knowledge Management and Organizational Memories, Kluwer, 2002.

[Wu et al, 2002] S.H. Wu, M.Y. Day, W.L. Hsu,Faq-Centered Organizational Memory. To appear in R. Dieng-Kuntz and N.Matta eds, Knowledge Management and Organisational Memories, Kluwer, 2002.

[Watson, 2002] I. Watson,A Knowledge Management Initiative By Uk Local Government. To appear in R. Dieng-Kuntz and N.Matta eds, Knowledge Management and Organizational Memories, Kluwer, 2002.

Information Retrieval

[Bamshad, 2001], Bamshad Mobasher, “ARCH: An Adaptive Agent for Retrieval Based on Concept Hierarchies”, SemanticWeb Mining Workshop at ECML/PKDD-2001, September 3, 2001, Freiburg, Germany.

[Cohen, 1999], Cohen, P. R., Chaudhri, V., Pease, A., and Schrag, R. “Does Prior Knowledge Facilitate the Development ofKnowledge-based Systems ?”, Proceedings of The Sixteenth National Conference on Artificial Intelligence (AAAI-99).http://projects.teknowledge.com/HPKB/Publications.html

[Crow, 2001], L. Crow, N. Shadbolt “Extracting focused knowledge from the semantic web”, Int. J. Human-Computer Studies54, pp 155-184, 2001.

[Gomez-Perez, 1999], Gomez-Perez, A. “Evaluation of Taxonomic Knowledge in Ontologies and Knowledge Bases”,Proceedings of KAW'99. http://sern.ucalgary.ca/KSI/KAW/KAW99/papers.html

[Leroy, 2000], G. Leroy, K. M. Tolle, H. Chen, “Customizable and Ontology-Enhanced Medical Information RetrievalInterfaces”, Management Information Systems Department, University of Arizona, USA.http://ai.bpa.arizona.edu/go/intranet/papers/Customizable-00.htm, 2000.

[Mc Ilraith, 2001], S.A. Mc Ilraith, T.C Son, H. Zeng “Semantic Web Services”, IEEE Intelligent Systems, pp. 46-53, 2001

[Menzies, 1998], Menzies, T. “Evaluation Issues with Critical Success Metrics”, Proceedings of KAW'98.http://ksi.cpsc.ucalgary.ca/KAW/KAW98/KAW98Proc.html, 1998.

[Nick, 1999], Nick M., Althoff, K., Tautz, C. “Facilitating the Practical Evaluation of Knowledge-Based Systems andOrganizational Memories Using the Goal-Question-Metric Technique”, Proceedings of KAW'99.http://sern.ucalgary.ca/KSI/KAW/KAW99/papers.html, 1999.

[Qi Li et al 2001], Qi Li, Philip Shilane, Natalya Fridman Noy, M.A. Musen “Ontology Acquisition from on-line KnowledgeSources”, AMIA Inc. pp. 497 – 501, 2001.

[Ricardo et al. 1999], Ricardo Baeza-Yates, Berthier Ribeiro-Neto "Modern Information Retrieval ", (Chapter 3 "RetrievalEvaluation") ACM Press, New York, Addison-Wesley, 1999


80/100

LinksSARI : http://www.ercim.org/publication/Ercim_News/enw35/taveter.htmlMCM: http://ai.bpa.arizona.edu/go/intranet/papers/Customizable-00.htm

E-business, E-commerce[E-work and E-commerce 2001] "Novel solutions and practices for a global networked economy", e.2001, 17-20 October 2001,Venice, Italy. Edited by Brian Stanford-Smith and Enrica Chiozza, ISBN: 1 58603 205 4.URL: http://www.ebew.net/programme.htm

[Chappell, 2000], C. Chappell, S. Feindt "Analysis of E-Commerce prcatice in SMEs", Communications and Strategies, N°37, 1sttrimester 2000

[Feindt et al. 2001], Sylvie Feindt et al., "Critical Success Factors for eBusiness for SMEs",in [E-work and E-commerce 2001]

[Dubosson-Torbay et Al., 2001], M. Dubosson-Torbay, A. Osterwalder, Y. Pigneur “eBusiness Model Design, Classification andMeasurements”, Thunderbird International Business Review, 2001.

Best Practices, LLC http://www.best-in-class.com

Portals for Web Community

[Staab S; et al. 2001] S. Staab, H.-P. Schnurr, R. Studer, and Y. Sure. "Knowledge processes and ontologies". IEEE IntelligentSystems, 16(1):26–34, 2001.

[Schreiber G. et al. 1999 ] G. Schreiber, H. Akkermans, A. Anjewierden, R. de Hoog, N. Shadbolt, W. Van de Velde, and B.Wielinga. Knowledge Engineering and Management — "The CommonKADS Methodology". The MIT Press, Cambridge,Massachusetts; London, England, 1999.

[Gomez-Pérèz A. 1996] A. Gomez-Perez. "A framework to verify knowledge sharing technology". Expert Systems withApplication, 11(4):519–529, 1996.

[Arpírez et al, 2001] Arpírez J.C., Corcho O., Fernández-López M., Gómez-Pérez A., “WebODE: a Workbench for Ontological

Engineering”, First International Conference on Knowledge Capture (K-CAP’01), Victoria, Canada, 2001.

[Bernaras et al, 1996] Bernaras A., Laresgoiti I., Corera J.,“ Building and Reusing Ontologies for Electrical Network

Applications”, Proceedings of the European Conference on Artificial Intelligence (ECAI’96). ECAI 96. Publisher: John Wiley &

Sons, pp298-302, 1996.

[Fernández-López et al., 1999] Fernández-López M., Gómez-Pérez A., Pazos-Sierra A., Pazos-Sierra J., “Building a Chemical

Ontology Using METHONTOLOGY and the Ontology Design Environment”, IEEE Intelligent Systems & theirapplications,

pp37-46, 1999.

[Fernández-López, 1999] Fernández López M., "Overview of methodologies for building ontologies", Ontologies and Problem-

Solving Methods: Lessons Learned and Future Trends, International Joint Conference on Artificial Intelligence, Stockholm, pp4-1

to 4-13, 1999.

[Grüninger et al., 1995] Grüninger M., Fox M.S., “Methodology for the design and evaluation of ontologies”, Workshop on Basic

Ontological Issues in Knowledge Sharing, Montreal, 1995.

[IEEE, 1996]IEEE Standard for Developing Software Life Cycle Processes. Std. 1074-1995. IEEE Computer Society. New York,

1996.

[Uschold et al., 1995] Uschold M., King M., “Towards a Methodology for Building Ontologies”, Workshop held in conjunction

with IJCAI on Basic Ontological Issues in Knowledge Sharing, 1995.


81/100

6 Tools and Methodology for Ontology-based Applications

6.1 Tools for ontology-based applications

While still not a widely spread commercial practice, the development and deployment of ontology-basedapplications can be supported by a variety of tools that currently exist on the market.

This section presents a list of some representative examples of tools that support various facets of anontology-based IR application. Although our main focus is information retrieval, we include in this list anumber of tools that support various processes/aspects that are closely related to IR. Tools are groupedinto the following categories10:

ÿ� Ontology related inference engines: tools in this category are concerned with the creation,maintenance and usually validation of ontologies. Usually they will contain an inference engine(based on some description logic) that will check for inconsistencies and also be able to dealwith queries concerning the ontology itself e.g. answering parent-child relationships, providinginheritance mechanisms etc.

ÿ� Topic maps: tools in this category are concerned with the organization of concepts in clustersand their visualization using 3D contour maps as a metaphor. Clusters are depicted as hills, theimportance of which is represented by the height and size of the hill. Tools in this category arehybrids combining both the organization and the visualization of content.

ÿ� Content management: tools in this category claim to cover a broad spectrum of ontologymanagement tasks such as modeling, categorization, integration, and indexing. They usuallyoffer a suite of tools for the creation and validation of ontology as well as for accessingorganizing and storing content on the basis of the ontology. Typical examples are Semiomap andXcellerant.

ÿ� Information retrieval : tools in this category focus on the extraction process itself although bynecessity they contain indexing and/or categorization capabilities as well. Inxight is a typicalexample here that does extraction and web content navigation.

ÿ� Information visualization : tools in this category are concerned with the presentation ofinformation in a user friendly, intuitive manner. Methods employed utilize some kind of a 2 or3D graph/network (e.g. hyperbolic trees) with various effects employed in order to representcontext, concept in focus and related entities/links while at the same time avoiding informationoverload. Typical examples here are TheBrain and Thinkmap, which employ a ‘neuronal’ typeof representation to depict concepts.

In the following, table 6.1 overleaf presents for each of the above groups a number of typical commercialproducts together with a link to the company’s site and a brief label of their exact nature.

10 The mentioning of commercial products in this section does not imply their endorsement by the authorsin any way. Products are selected randomly and discussed only as typical examples of their category.


82/100

COMPANY PRODUCT PRODUCT CATEGORY

ONTOLOGY RELATED INFERENCE ENGINESCycorp

www.cyc.comCyc Knowledge Server

Multi-contextual knowledge base /inference engine

Ontoprisewww.ontoprise.de

Ontobroker Inference Middleware

Network Inferencewww.networkinference.com

Cerebra Inference engine and tools

TOPIC MAPSEmpolis

www.empolis.co.ukK42 Topic Map Server

Infoloomwww.infoloom.com

Topic Map Loom Topic Map editor

Mondecawww.mondeca.com

SeveralTopic Maps to improve Content

Management

Ontopiawww.ontopia.net

Topic Map EngineTopic Map Navigator

Topic Map Client and Server

CONTENT MANAGEMENT

Applied Semanticswww.appliedsemantics.com

CircaOntology-based automatic

categorization

Unicorn Solutionswww.unicorn.com

Unicorn CoherenceOntology modelling and data

integration

DigitalOwlwww.digitalowl.com

KineticEdge Content Management / Publishing

H5 Technologieswww.h5technologies.com

H5 AtlasH5 AutoTagger

H5 Syndica

Content categorization

Semiowww.semio.com

SemioMap Content Categorization and Indexing

Eprisewww.eprise.com

Participant Server Content Management

Epigraphwww.epigraph.com

XcellerantContent Management / Ontology

Management

Forward look inc.www.forwardlook.com

ContextStreams Data Asset Management

Global Wisdomwww.globalwisdom.org

Bravo engineOntology Construction / Dynamic

Knowledge Engine

Persistwww.persistag.com

Semantic Base Knowledge Management System

Profiumwww.profium.com

Smart Information Router (SIR)Semantic Content Management

based on RDF

Voquettewww.taalee.com

Semantic EngineWorld Model

Knowledge-based ContentManagement

INFORMATION RETRIEVAL

Inxightwww.inxight.com

ThingFinder ServerStar Tree Viewer

Content extractionWeb content navigation

Mohominewww.mohomine.com

severalInformation extraction and

classification

INFORMATION VISUALIZATION


83/100

AIdministratorwww.aidministrator.nl

SesameSpectacle

Ontology-based informationpresentation

Thinkmapwww.thinkmap.com

ThinkmapInformation visualization and

organization

TheBrain.comwww.thebrain.com

TheBrain Information organizer

Table 6.1 - List of tools for ontology related applications

6.2 Ontology engineering11

Ontologies aim at capturing domain knowledge in a generic way and provide a commonly agreedunderstanding of a domain, which may be reused and shared across applications and groups. Ontologiestypically consist of definitions of concepts, relations and axioms. Until a few years ago the building ofontologies was done in a ratherad hoc fashion. Meanwhile there have been some few, but seminalproposals for guiding the ontology development process (cf.6.1.2).In this chapter we describe a methodology for building an ontology-based application by introducingspecific guidelines for developing and maintaining the respective ontology. Special emphasis is put on astepwise construction and evaluation of the ontology.

Figure 6.1 - Steps of the Ontology Building methodology

6.2.1 Methodology

Kick-off phase. The first step to actually engineer ontologies is to capture requirements in an OntologyRequirements Specification Document (“ORSD”) describing what ontology should support and sketchingthe planned area of the ontology application. It should guide an ontology engineer to decide aboutinclusion, exclusion and the hierarchical structure of concepts in the ontology. In this early stage oneshould look for already developed and potentially reusable ontologies. In summary, it should clearlydescribe the information shown in table 6.2.

11 The content is extracted from [Sure Y. 2002]


84/100

Through analysis of the available knowledge sources, “baseline ontology” is gathered,i.e. a draft versioncontaining few but seminal elements of ontology. Typically the most important concepts and relations areidentified on an informal level. A very important knowledge source (also for the later phases) is domainexperts. There exist several possibilities to capture knowledge from domain experts; we focus on theusage of informal competency questionnaires as proposed by [Bernaras A. 1996]. They consist ofpossible queries (“competency questions”) to the system, indicating the scope and content of theontology.

ORSD1.Domain and goal of the ontologyORSD2.Design guidelines to ensure a consistent development (e.g.naming conventions)ORSD3. Available knowledge sources (e.g. domain experts, reusable ontologies, organization charts,

business plans, dictionaries, index lists, db-schemas etc.)ORSD4.Potential users and use casesORSD5.Applications supported by the ontology

Table 6.2 - Content of the ORSD

Refinement phase. The goal of the refinement phase is to produce a mature and application-oriented“target ontology” according to the specification given by the kickoff phase. This phase is divided intodifferent sub phases shown in table 6.3.

R1. A knowledge elicitation process with domain experts based on the initial input from the kickoffphase. This serves as input for further expansion and refinement of the baseline ontology. Typicallyaxioms are identified and modeled in this phase. This is closely linked to the next step – the effects ofaxioms might depend on the selection of the representation language.

R2. A formalization phase to transfer the ontology into the “target ontology” expressed in formalrepresentation languages like DAML+OIL [DAML 2001]. The representation language is chosenaccording to the specific requirements of the envisaged application.

Table 6.3 - Two sub phases of the refinement phase

This phase is closely linked to the evaluation phase. If the analysis of the ontology in the evaluation phaseshows gaps or misconceptions, the ontology engineer takes these results as an input for the refinementphase. It might be necessary to perform several iterative steps.

Evaluation phase. The evaluation phase serves as a proof for the usefulness of developed ontologies andtheir associated software environment. In a first step, the ontology engineer checks, whether the targetontology fulfils the ontology requirements specification document and whether the ontology supports or“answers” the competency questions analyzed in the kick-off phase of the project. In a second step, theontology is tested in the target application environment. Feedback from beta users may be a valuableinput for further refinement of the ontology.

A valuable input may be as well the usage patterns of the ontology. The prototype system has to trackthe ways users navigate or search for concepts and relations. With such an “ontology log file analysis” wemay trace what areas of the ontology are often “used” and others which were not navigated. Asmentioned before, this phase is closely linked to the refinement phase and an ontology engineer may needto perform several cycles until the target ontology reaches the envisaged level— the roll out of the targetontology embedded into the ontology-based application finishes the evaluation phase.

Maintenance phase. In the real world things are changing — and so do the specifications for ontologies.To reflect these changes ontologies have to be maintained frequently like other parts of software, too. Westress that the maintenance of ontologies is primarily an organizational process. There must be strict rulesfor the update-insert-delete processes within ontologies. Most important is to clarifywho is responsiblefor maintenance andhow it is performed.E.g. is a single person or a consortium responsible for the


85/100

maintenance process? In which time intervals is the ontology maintained? We recommend that theontology engineer gathers changes to the ontology and initiates the switchover to a new version of theontology after thoroughly testing possible effects to the application,viz. performing additional cyclicrefinement and evaluation phases. Similar to the refinement phase, feedback from users may be a valuableinput for identifying the changes needed. Maintenance should accompany ontologies as long as they areon duty.

6.2.2 Related WorkWe here give an overview of existing methodologies for ontology development and show which of theirideas are adopted and expanded in our methodology.

Skeletal Methodology.This methodology is based on the experience of building the Enterprise Ontology(cf. [Ushold et al. 1995]), which includes a set of ontologies for enterprise modeling. The guidelines fordeveloping ontologies start with identifying the purpose of ontology and then concentrate on the buildingof ontologies which is broken down into the steps ontology capture, coding, evaluation anddocumentation. A disadvantage of this methodology is that it does not precisely describe the techniquesfor performing the different activities. For example, it remains unclear, how the key concepts andrelationships should be acquired. Only a very vague guideline is given.

We catch up the idea of competency questions and expand their usage. We not only propose to usethem for evaluation of the system, but also for finding relevant lexical entries like concepts, relations etc..

KACTUS . The approach of [Bernaras A. 1996] was developed within the Esprit KACTUS project. Oneof the objectives of this project was to investigate the feasibility of knowledge reuse in complex technicalsystems and the role of ontologies to support it. The methodology recommends an application drivendevelopment of ontologies. So, every time an application is assembled, the ontology that represents theknowledge required for the application is built. Three steps have to be taken every time an ontology-basedapplication is assembled:

1. Specification of the application. Provide an application context and a view of the componentsthat the application tries to model.

2. Preliminary design. Based on relevant top-level ontological categories create a first draft wherethe list of terms and application specific tasks developed during the previous phase is used asinput for obtaining several views of the global model in accordance with the top-levelontological categories determined. Search for existing ontologies which may be refined andextended for use in the new application.

3. Ontology refinement and structuring. Structure and refine the model in order to arrive at adefinitive design.

The methodology offers very little detail and does not recommend particular techniques to support thedevelopment steps. Also, documentation, evaluation and maintenance processes are missing [7]. Ingeneral we agree with the general idea of application driven ontology development and in particular withrefinement and structuring, which is reflected by our proposal of the ontology development process.

Methontology.The Methontology framework from [Gomez A. 1998] includes:

1. The identification of the ontology development process, which refers to which tasks (planning,control, specification, knowledge acquisition, conceptualisation, integration, implementation,evaluation, documentation, configuration management) one should carry out, when buildingontologies.

2. The identification of stages through which ontology passes during its lifetime.3. The steps to be taken to perform each activity, supporting techniques and evaluation steps.4. Setting up an ontology requirements specification document (ORSD) to capture requirements for

ontology similar to a software specification.


86/100

The methodology offers detailed support in development-oriented activities except formalization andmaintenance and describes project management activities. We adopted the general idea of an ontologyrequirements specification document (ORSD), but modified and extended the presented version by ourown needs.

6.3 Ontology building from text12

One promising approach for establishing an ontology and acquire knowledge is to incorporate resultsfrom disciplines like linguistics. Researchers in terminology for example are interested in organizingdomains from a conceptual point of view from the analysis of terms used to name concepts in texts. Onthe other hand, an ontology is based on the definition of a structured and formalized set of concepts, and agreat part of it comes from text analysis, such as transcript of interviews, and technical documentation. Insuch cases, the theory of a domain can only be found by reaching concepts from terms.For several years, some researchers in terminology have identified a parallel between terminology as apractical discipline and artificial intelligence, in particular knowledge engineering. From a knowledgeengineering point of view, we notice two trends. One trend is to propose to elicit knowledge by usingautomatic processing tools, widely used in linguistics. Another one is to establish a synergy betweenresearch works in artificial intelligence and in linguistics, by means of terminology. An overview of thesedevelopments is given below.

Natural language processing tools may help to support modeling from texts in two ways. First, they canhelp to find the terms of a domain [Bou94], [BGG96] [OFR96]. Existing terminologies or thesauri maybe reused and increased or new ones may be created. Second, they can help to structure a terminologicalbase by identifying relations between concepts [Jou95] [JME95] [Gar97].Three steps are necessary to find the terms of a domain. At the beginning, nominal groups are isolatedfrom a corpus considered as being representative of the studied domain. Then, those that can't be chosenas terms because of morphological or semantic characteristics are eliminated. Finally, the nominalsequences that will be retained as terms are chosen. Usually, this last step requires a human expertise.Identifying relations between concepts is composed of three steps too. The first one identifies the co-occurrences of terms. Two terms are co-occurrent if they both appear in a given text window which maybe defined in several ways: a number of words, a documentary segmentation (entire document, section), asyntactic cutting of sentences, ... The second step computes a similarity between terms with respect tocontexts they share. Then, the third step can determine the terms that are semantically related. In mostcases, identified relations are the following: semantic proximity, meronimy, causal or more specificrelations.

Some researchers have focused on trying to benefit from approaches from both linguistics and knowledgeengineering. They have studied mutual contributions, and their work has led them to elaborate the conceptof Terminological Knowledge Base (TKB). This concept was first defined by Ingrid Meyer [SMe91][MSB+92].

Building a TKB is seen as an intermediate model that helps toward the construction of a formal ontology.A TKB is a computer structure that contains conceptual data, represented in a network of domainconcepts, but also linguistic data on the terms used to name the concepts. Thus a TKB contains threelevels of entities: term, concept and text. It is structured by using three kinds of links. Relations betweenterm and concept allow synonymy and paronimy to be considered. Relations between concepts composethe network of domain concepts. Relations between term and/or concept and text allow normalizationchoices to be justified or knowledge base to be documented. A TKB is interesting to build a KBS,especially because it gathers some linguistic information on terms used to name concepts on. This canenhance communication between experts, knowledge engineers and end-users, or be a great help for the

12 Contribution from Univ. d’Orsay, Paris Sud, LRI (Chantal Reynaud) as in FIPA Part 12 Ontology services (Annex B Guidelines

to define a new Ontology under editorship of A. Léger FIPA 1998]


87/100

knowledge engineer to choose the names of the concepts in the system. Nevertheless, if most researchersagree with its structure, problems still remain today about genericity and also about the construction andthe exploitation of the corpus, which is very important in the construction of the TKB because it is thereference from which modeling choices will be justified. Current research continues in these directions.

6.4 References

[Ushold M. 1995] M. Uschold and M. King. "Towards a Methodology for Building Ontologies". In Workshop on Basic

Ontological Issues in Knowledge Sharing, held in conjunction with IJCAI-95, Montreal, Canada, 1995.

[Gomez 1996] A. Gomez-Perez. "A Framework to Verify Knowledge Sharing Technology". Expert Systems with Application,

11(4):519– 529, 1996.

[Ushold M. 1996] M. Uschold & M. Grüninger, "Ontologies: Principles, methods and applications" Knowledge Sharing and

Review, 11(2)

[Bernaras A. et al. 1996] A. Bernaras, I. Laresgoiti, and J. Corera. "Building and Reusing Ontologies for Electrical Network

Applications". In Proceedings of the European Conference on Artificial Intelligence ECAI-96, 1996.

[Gomez-Perez A. 1998] Assuncion Gomez-Pérez, "Methontology" « Knowledge Sharing and Reuse »,Laboratorio de

Intelligencia Artificial, Facultad de informatica, Universidad Politécnica de Madrid.

[D. O'Leary 1998] D. O’Leary. "Using AI in knowledge management: Knowledge bases and ontologies". IEEE Intelligent

Systems, 13(3):34– 39, May/June 1998.

[Lopez F 1999] F. L´opez. "Overview of methodologies for building ontologies". In Proceedings of the IJCAI-99 Workshop on

Ontologies and Problem-Solving Methods: Lessons Learned and Future Trends. CEUR Publications, 1999.

[Sure Y. 1999] York Sure, Rudi Studer, "On-To-Knowledge Methodology - Employed and Evaluated Version", On-To-

Knowledge Project IST-1999-10132, pp. 23-30

[Staab S 2001] S. Staab, H.-P. Schnurr, R. Studer, and Y. Sure. "Knowledge processes and ontologies". IEEE Intelligent Systems,

16(1):26– 34, 2001.

[Sure Y. 2001 Y. Sure and R. Studer. "On-To-Knowledge Methodology — Evaluated and Employed Version". On-To-

Knowledge deliverable D-16, Institute AIFB, University of Karlsruhe, 2001.

[Sure Y. 2002] York Sure, "A Tool-supported Methodology for Ontology-based Knowledge Management", submitted to ISMIS

2002, Methodologies for Intelligent Systems

[DAML 2001] DAML+OIL . http://www.daml.org/2001/03/daml+oil-index

[BCo95] Bourigault D., Condamines A., "Réflexions autour du concept de base de connaissances Terminologiques", Dans les

actes des journées nationales du PRC-IA, Nancy, 1995.

[Bou94] Bourigault D., "LEXTER, un logiciel d'extraction de terminologie. Application à l'acquisition des connaissances à

partir de textes", Thèse de l'Ecole des Hautes Etudes en Sciences Sociales (Paris).

[BGG96] Bourigault D., Gonzalez-Mullier I., Gros C., "LEXTER, a natural Language Processing Tool for Terminology

Extraction", actes de EURALEX'96 (Göteborg).

[Gar97] GARCIA D., "COATIS, an NLP System to Locate Expressions of Ations Connected by Causality Links", in Proc.

10th European Workshop, EKAW'97, San Feliu de Guixols, Catalonia, Spain, October 97, LNAI 1319, pp. 347-352, 1997.

[Jou95] Jouis Ch., "SEEK, un logiciel d'acquisition des connaissances utilisant un savoir linguistique sans employer de

connaissances sur le monde externe", Actes des 6èmes Journées Acquisition et Validation (JAVA'95), Grenoble, pp.159-172,

1995.

[JME95] Jouis Ch., Mustafa-Elhadi W., "Conceptual Modeling of database Schema using linguitic knowledge. Application to

terminological Knowledge bases", First Workshop on Application of Natural language to Databases (NLDB'95), Versailles, Juin

95, pp. 103-118, 1995.

[MSB+92] Meyer I., Skuce D., Bowker L., Eck K., "Toward a new generation of terminological resources: an experiment in

building a terminological knowledge base. In Proc. 14th International Conference on Computational Linguistics. Nantes. pp. 956-

960, 1992.

[OFR96] Oueslati R., Frath P., Rousselot F., "Term identification and Knowledge Extraction", International Conference on

Applied Natural Language and Artificial Intelligence. Montreal. Juin 96.


88/100

7 Conclusions

D2.2 focuses on the identification of the most suitable techniques that may be applied to each cluster ofontology-based applications. It provides guidelines to assist an organisation on what techniques could beapplied for a given application. In this deliverable D2.2, most representative applications and guidelinesare precisely defined for both application domains: Information Retrieval, Enterprise Portals andKnowledge management. For these application domains, specific and generic criteria for evaluation arepretty well identified and given. On the contrary, it was more difficult, while rendering more mature thecontent dedicated to e-Commerce and Portals & Web Communities, to identify reference evaluationmethodology, reference evaluation criteria or metrics and specific ontology enhanced criteria. However,taking some distance from the work realized on Information Retrieval and Enterprise Portal & KM, wehave now the feeling that some results could be reused and may be applicable for e-Commerce andPortals & Web Communities. This work will be continued in the next deliverable for these two lastdomains of application.

Moreover, regarding and analyzing the results we have obtained and presented in the deliverable D2.2,we have now the feeling that ontology-based application and/or service evaluation could be seen under aweb-services angle, for which the view of end-users becomes one of the key evaluation criteria to assesson the added-value of the introduction of ontology-based applications in the organization's business area.Such assistance and assessment on this added value will be given in the next deliverable D2.3. D2.3 willprovide guidelines to assist an organization on why it should use ontologies.


89/100

8 References

Evaluations[Rombach, 1991], Rombach H.D. ,“Practical benefits of goal-oriented measurement”, in Fenton and Littlewood editors,

Software Reliability and Metrics, pages 217-235. Elsevier Applied Science, London.

[Grüninger et al.] Michael Grüninger, Mark S. Fox,"Methodology for the Design and Evaluation of Ontologies",Proceedings of

IJCAI'95, Workshop on Basic Ontological Issues in Knowledge Sharing, April 13, 1995

[Cohen, 1995]Empirical Methods for Artificial Intelligence , MIT Press

[Briand 1996], Briand L.C., Differding C.M., Rombach H.D. ,“Practical guidelines for measurement-based process

improvement”, Software process, 2(4): pages 253-280

[Menzies, 1997] T. Menzies, "Evaluation Issues for Problem Solving Methods" , Banff KA workshop 1998

[Menzies, 1998] T. Menzies, "Evaluation Issues with Critical Success Metrics", Proceedings of KAW’98

[Menzies, 1998] T. Menzies, "Requirements for good measurements", http://www.cse.unsw.edu.au/~timm/pub/eval/cautions.html



[Jasper et al. 1999] R. Jasper and M. Uschold, "A framework for Understanding and Classifying Ontology Applications ",

Proceedings of KAW'99

[Benjamin et al. 1999],Asuncion Gomez Perez, V. Richard Benjamins, "Overview of Knowledge, Sharing and Reuse

Components: Ontologies and Problem-Solving Methods". In Proceedings of the IJCAI-99 workshop on Ontologies and Problem-

Solving Methods (KRR5) Stockholm, Sweden, August 2, 1999. pp. 1.1--1.15

[Sure et al. 1999] York Sure, Rudi Studer, "On-To-Knowledge Methodology - Employed and Evaluated Version", On-To-

Knowledge Project IST-1999-10132

[Rausher 1999] J. Rauscher, S. Marc, "A la conquête de la Silicon Valley" , Editions d’Organisation, Paris, ISBN: 2-7081-2282-7

[Julieanne van Zyl, Dan Corbett, 2000 a]"Population of a Framework for Understanding and Classifying Ontology

Applications" , ECAI'2OOO

[Julieanne van Zyl, Dan Corbett, 2000 b]"A Framework for Comparing the use of a Linguistic Ontology in an Application" ,

ECAI'2000



Knowledge Management

[Staab et al., 2001], Steffen Staab, Rudi Studer, Hans-Peter Schnurr, York Sure,“Knowledge Management : Knowledge

Processes and Ontologies”; IEEE Intelligent Systems Journal, Jan. Feb. 2001.

[Dieng et al., 1999], R. Dieng, O. Corby, A. Giboin, and M. Ribiere.Methods and tools for corporate knowledge management.


[Kuehn et al., 1997], O. Kuehn and A. Abecker.Corporate memories for knowledge memories in industrial practice: Prospects

and challenges. Journal of Universal Computer Science, 3(8), August 1997.

[Schreiber et al., 1999], G. Schreiber, H. Akkermans, A. Anjewierden, R. de Hoog, N. Shadbolt, W. Van de Velde, and B. Wielinga.

Knowledge Engineering and Management — The CommonKADS Methodology. The MIT Press, Cambridge, Massachusetts;

London, England, 1999.

[Dyson S. IBM, 2001], S. Dyson, IBM UK, «B2E Mobile e-Business : Driver, Passenger or Spectator ?», in [E-work and E-

commerce 2001].

[D. fensel, 2001] D. Fensel: "Ontologies: Silver Bullet for Knowledge Management and Electronic Commerce". Springer-

Verlag, Berlin, 2001.

[D. Fensel et al. 2002] D. Fensel, J. Hendler, H. Lieberman, and W. Wahlster (eds.): "Semantic Web Technology", MIT Press,

Boston, to appear.

[D. Fensel, 1998], D. Fensel, S. Decker, M. Erdmann, and R. Studer: "Ontobroker in a Nutshell" (short paper). In C. Nikolaou et

al. (eds.), Research and Advanced Technology for Digital Libraries,Lecture Notes in Computer Science, LNCS 1513, Springer-

Verlag Berlin, 1998.

[O'Leary, et al. 2001] Dan E. O’Leary and Rudi Studer (eds.): "Knowledge Management", IEEE Intelligent Systems 16 (1),

January/February 2001 (special issue).


90/100

[Staab S. et al. 2001] Steffen Staab and Alexander Maedche, "Knowledge Portals — Ontologies at Work". AI Magazine21(2),

Summer 2001.

Steffen Staab, Rudi Studer, Hans-Peter Schnurr, York Sure“Knowledge Management : Knowledge Processes and Ontologies”;,

IEEE Intelligent Systems Journal, Jan. Feb. 2001.


Ontology Building and Checking[Ushold M. 1995] M. Uschold and M. King. "Towards a Methodology for Building Ontologies". In Workshop on Basic

Ontological Issues in Knowledge Sharing, held in conjunction with IJCAI-95, Montreal, Canada, 1995.

[Gomez 1996] A. Gomez-Perez. "A Framework to Verify Knowledge Sharing Technology". Expert Systems with Application,

11(4):519– 529, 1996.

[Ushold M. 1996] M. Uschold & M. Grüninger, "Ontologies: Principles, methods and applications" Knowledge Sharing and

Review, 11(2)

[Bernaras A. et al. 1996] A. Bernaras, I. Laresgoiti, and J. Corera. "Building and Reusing Ontologies for Electrical Network

Applications". In Proceedings of the European Conference on Artificial Intelligence ECAI-96, 1996.

[Borst P. et al. 1997] Borst P. and Akkermans H.« Engineering ontologies », Special issue : Using explicit ontologies in

knowledge-based system development, HCS, Vol. 46, Number 2/3, February/March 1997, pp. 365-406.

[Gomez-Perez A. 1998] Assuncion Gomez-Pérez, "Methontology" « Knowledge Sharing and Reuse »,Laboratorio de

Intelligencia Artificial, Facultad de informatica, Universidad Politécnica de Madrid.

[D. O'Leary 1998] D. O’Leary. "Using AI in knowledge management: Knowledge bases and ontologies". IEEE Intelligent

Systems, 13(3):34– 39, May/June 1998.

[Lopez F 1999] F. L´opez. "Overview of methodologies for building ontologies". In Proceedings of the IJCAI-99 Workshop on

Ontologies and Problem-Solving Methods: Lessons Learned and Future Trends. CEUR Publications, 1999.

[Sure Y. 1999] York Sure, Rudi Studer, "On-To-Knowledge Methodology - Employed and Evaluated Version", On-To-

Knowledge Project IST-1999-10132, pp. 23-30

[Staab S 2001] S. Staab, H.-P. Schnurr, R. Studer, and Y. Sure. "Knowledge processes and ontologies". IEEE Intelligent Systems,

16(1):26– 34, 2001.

[Sure Y. 2001 Y. Sure and R. Studer. "On-To-Knowledge Methodology — Evaluated and Employed Version". On-To-

Knowledge deliverable D-16, Institute AIFB, University of Karlsruhe, 2001.

[Sure Y. 2002] York Sure, "A Tool-supported Methodology for Ontology-based Knowledge Management", submitted to ISMIS

2002, Methodologies for Intelligent Systems

[DAML 2001] DAML+OIL . http://www.daml.org/2001/03/daml+oil-index

Natural Language based Knowledge acquisition references

[BCo95] Bourigault D., Condamines A., "Réflexions autour du concept de base de connaissances Terminologiques", Dans les

actes des journées nationales du PRC-IA, Nancy, 1995.

[Bou94] Bourigault D., "LEXTER, un logiciel d'extraction de terminologie. Application à l'acquisition des connaissances à

partir de textes", Thèse de l'Ecole des Hautes Etudes en Sciences Sociales (Paris).

[BGG96] Bourigault D., Gonzalez-Mullier I., Gros C., "LEXTER, a natural Language Processing Tool for Terminology

Extraction", actes de EURALEX'96 (Göteborg).

[Gar97] GARCIA D., "COATIS, an NLP System to Locate Expressions of Ations Connected by Causality Links", in Proc.

10th European Workshop, EKAW'97, San Feliu de Guixols, Catalonia, Spain, October 97, LNAI 1319, pp. 347-352, 1997.

[Jou95] Jouis Ch., "SEEK, un logiciel d'acquisition des connaissances utilisant un savoir linguistique sans employer de

connaissances sur le monde externe", Actes des 6èmes Journées Acquisition et Validation (JAVA'95), Grenoble, pp.159-172,

1995.

[JME95] Jouis Ch., Mustafa-Elhadi W., "Conceptual Modeling of database Schema using linguitic knowledge. Application to

terminological Knowledge bases", First Workshop on Application of Natural language to Databases (NLDB'95), Versailles, Juin

95, pp. 103-118, 1995.


91/100

[MSB+92] Meyer I., Skuce D., Bowker L., Eck K., "Toward a new generation of terminological resources: an experiment in

building a terminological knowledge base. In Proc. 14th International Conference on Computational Linguistics. Nantes. pp. 956-

960, 1992.

[OFR96] Oueslati R., Frath P., Rousselot F., "Term identification and Knowledge Extraction", International Conference on

Applied Natural Language and Artificial Intelligence. Montreal. Juin 96

[Skuce D. et al 1998] Skuce D., Meyer I., "Terminology and knowledge acquisition: exploring a symbiotic relationship". In

Proc. 6th Knowledge Acquisition for Knowledge-Based System Workshop, Banff, pp. 29/1-29/21.

[Houssem A. 1998] Houssem Assadi, "Construction of a regional ontology from text and its use within a documentary

system", FOIS’98, pp. 236-249, Trento, June 98.

E-Commerce and E-Business[E-work and E-commerce 2001] "Novel solutions and practices for a global networked economy", e.2001, 17-20 October 2001,

Venice, Italy. Edited by Brian Stanford-Smith and Enrica Chiozza, ISBN: 1 58603 205 4.URL:

http://www.ebew.net/programme.htm

[eWork 2001], "Status Report on New Ways to Work in the Knowledge Economy", IST document

URL: http://www.eto.org.uk

[MKBEEM 2000], http://www.mkbeem.com/

[MULECO 200], http://www.cenorm.be/isss/Workshop/ec/MULECO/Documents/Muleco_Documents.htm

[SmartEC, 2000],http://www.telecom.ntua.gr/smartec/

[Chappell, 2000], C. Chappell, S. Feindt "Analysis of E-Commerce prcatice in SMEs", Communications and Strategies, N°37, 1st

trimester 2000

[Feindt et al. 2001], Sylvie Feindt et al., "Critical Success Factors for eBusiness for SMEs",in [E-work and E-commerce 2001]

Information retrieval[Guarino et al., 1999], Nicola Guarino, Claudio Masolo, Guido Vetere, "OntoSeek: Content-Based Access to the Web", IEEE

Intelligent Systems, 1999.

[Stuart Aitken 2000] Stuart Aitken, Sandy Reid,"Evaluation of an Ontology-Based Information retrieval Tool" , ECAI'00,

Applications of Ontologies and Problem-Solving Methods.

[Fensel et al., 2000], Dieter Fensel, Stefan Decker, Michael Erdmann, Rudi Studer, "Ontobroker: Ontology based Access to

Distributed and Semi-Structured Information"

[Qi Li et al 2001], Qi Li, Philip Shilane, Natalya Fridman Noy, M.A. Musen “Ontology Acquisition from on-line Knowledge

Sources”, AMIA Inc. pp. 497 - 501

[L. Crow 2001], L. Crow, N. Shadbolt “Extracting focused knowledge from the semantic web” Int. J. Human-Computer Studies

54, pp 155-184

[S. A. McIlraith 2001], S.A. McIlraith, T.C Son, H. Zeng “Semantic Web Services”. IEEE Intelligent Systems, pp. 46-53[Sure et al., 2002], York Sure, Michael Erdmann, Jürgen Angele, Steffen Staab, Rudi Studer, Dirk Wenke. OntoEdit: Collaborativeontology development for the semantic web. InProceedings of the ISWC 2002, June 9-12 2002, Sardinia, Italia., 2002.

SARI : http://www.ercim.org/publication/Ercim_News/enw35/taveter.html

MCM: http://ai.bpa.arizona.edu/go/intranet/papers/Customizable-00.htm

Portals for Web community[Stojanovic et al., 2001], N. Stojanovic, A. Maedche, S. Staab, R. Studer, and Y. Sure. "SEAL— A Framework for Developing

SEmantic PortALs" . In K-Cap 2001 - First International Conference on Knowledge Capture, Oct. 21-23, 2001, Victoria, B.C.,

Canada, 2001. to appear.

URL : http://ontobroker.semanticweb.org/ontos/aifb.html

[Saglio et al. 2002], J. M. Saglio, Tuang Anh Ta, "A Framework for Dynamic Exploration in Semantic Portals", KR2002 to appear

C-Web : Community Webs

URL: http://cweb.inria.fr/

[Staab S. et al. 2000] S. Staab, J. Angele, S. Decker, A. Hotho, A. Maedche, H-P. Schnurr, R. Studer, and Y. Sure. "AI for the web

— ontology-based community web portals". In AAAI 2000/IAAI 2000 - Proceedings of the 17th National Conference on Artificial

Intelligence and 12th Innovative Applications of Artificial Intelligence Conference. AAAI Press/MIT Press, 2000.


92/100

[Ontobroker, 1999] S. Decker, M. Erdmann, D. Fensel, and R. Studer. "Ontobroker: Ontology Based Access to Distributed and

Semi-Structured Information". In R. Meersman et al., editors,Database Semantics: Semantic Issues in Multimedia Systems, pages

351–369. Kluwer Academic Publisher, 1999.

[SHOE 2000] J. Heflin and J. Hendler. "Searching the web with SHOE". In Artificial Intelligence for Web Search. Papers from

the AAAI Workshop. WS-00-01, pages 35–40. AAAI Press, 2000.

[Staab S; et al. 2001] S. Staab, H.-P. Schnurr, R. Studer, and Y. Sure. "Knowledge processes and ontologies". IEEE Intelligent

Systems, 16(1):26–34, 2001.

[Schreiber G. et al. 1999 ] G. Schreiber, H. Akkermans, A. Anjewierden, R. de Hoog, N. Shadbolt, W. Van de Velde, and B.

Wielinga. Knowledge Engineering and Management — "The CommonKADS Methodology". The MIT Press, Cambridge,

Massachusetts; London, England, 1999.

[Gomez-Pérèz A. 1996] A. Gomez-Perez. "A framework to verify knowledge sharing technology". Expert Systems with

Application, 11(4):519–529, 1996.

General Web Sites

KA Workshops and Archives, http://ksi.cpsc.ucalgary.ca/KAW/

KAW'98 Proceedings,http:// ksi.cpsc.ucalgary.ca/KAW/KAW98/KAW98Proc.html

KAW'99 Proceedings,http://sern.ucalgary.ca/KSI/KAW/KAW99/papers.html

IETF Benchmarking methodology (bmwg); http://www.ietf.org/html.charters/bmwg-charter.html

Best Practices LLC,http://www.best-in-class.com/

ISMBC - Information Systems Management Benchmarking Consortium; http://www.ismbc.org/


SWWS Semantic Web Working Symposium, International Semantic Web Working Symposium, Stanford University, CA

USAAugust 1, 2001,http://www.semanticweb.org/SWWS/.


93/100

Annex 1

SEAL Description

8.1 HistoryThe origins of SEAL lie in Ontobroker[Decker, 1999], which was conceived for semantic searchof knowl-edge on the Web and also used for sharing knowledge on the Web[Benjamins, 1999],also taking advantage of the mediation capabilities of ontologies[Fensel, 2000]. It thendeveloped into an overarching framework for search and presentation offering access at a portalsite [Staab, 2000]. This concept was then transferred to further applications[Angele, 2000], [Sure,2000] and constitutes the technological basis for the portal of our institution (among others).Itnow combines the roles of information integration in order to provide data for the SemanticWeb and for a Peer-to-Peer network with presentation to human Web surfers.

8.2 Web Information IntegrationOne of the core challenges when building a data-intensive web site is the integration ofheterogeneous information on the WWW. The recent decade has seen a tremendous progress inmanaging semantically heterogeneous data sources[Fernandez, 2000][Wiederhold, 1997]. Thegeneral approach we pursue is to “lift” all the different input sources onto a common datamodel, in our case RDF. Additionally, an ontology acts as a semantic model for theheterogeneous input sources. As mentioned earlier and visualized in our conceptual architecturein figure 3.8, we consider different kinds ofdata sourcesof the Web as input: First of all, to alarge part the Web consists of static HTML pages, often semi-structured, including tables, lists,etc. We have developed an ontology-basedHTML wrapper that is based on a semi-supervisedannotation approach. Thus, based on a set of predefined manually annotated HTML pages, thestructure of new HTML pages is analyzed, compared with the annotated HTML pages andrelevant information is extracted from the HTML page. The HTML wrapper is currentlyextended to also deal with heterogeneous XML files. Second, we use an automatic XMLwrapping approach that has been introduced in[Erdmann, 2001]. The idea behind this wrappingapproach is that these XML documents refer to an DTD that has been generated from theontology. Therefore we automatically generate a mapping from XML to our data model so thatintegration comes for free. Third, data-intensive applications typically rely on relationaldatabases. A relational database wrapping approach[Stojanovic, 2002]maps relational databaseschemas onto ontologies that form the semantic basis for the RDF statements that areautomatically created from the relational database. Fourth, in an ideal case content providershave been registered and agreed to describe and enrich their content with RDF-based metadataaccording to a shared ontology. In this case, we may easily integrate the content automaticallyby executing anintegration process. If content providers have not been registered, but provideRDF-based metadata on their Web pages, we use ontology-focused metadata discovery andcrawling techniques to detect relevant RDF statements.

Our generic Web information integration architecture is extensible, as shown in figure 3.8. Inparticular, we are currently working on connecting and integrating data sources available viaenhancedPeer-2-Peer (P2P)networks. P2P applications for searching and exchanginginformation over the Web have become increasingly popular. TheEdutella approach buildsupon the RDF metadata standard aiming to provide an RDF-based metadata infrastructure forP2P applications, building on the recently announced JXTA framework.

It is important to mention that in our current architecture and implementation we mainly applystatic information integration building on a warehousing approach. Means fordynamicinformation integration are currently approached for Peer-2-Peer networks and within ourrelational database wrapper.


94/100

8.3 Web Site ManagementOne difficulty of community portals lies in integrating heterogeneous data sources. Each sourcemay be hosted by different community members or external parties and fulfils differentrequirements. Therefore typically all sources vary in structure and design. Community portalslike (in our case) the web site of our own institute require coherence in hosted information ondifferent levels. While the information integration aspect (see previous section) satisfies theneed for a coherent structure that is provided by the ontology we will now introduce variousfacilities for construction and maintenance of websites to offer coherent style and design. Eachfacility is illustrated by our conceptual architecture.Presentation view.Based on the integrated data in the warehouse we define user-dependentpresentation views. First, as a contribution to the Semantic Web, our architecture is dedicated tosatisfy the needs of software agents and produces machine understandable RDF. Second, werender HTML pages for human agents. Typicallyqueries for contentof the warehouse definepresentation views by selecting content, but alsoqueries for schemamight be used, e.g. to labeltable headers.Input view. To maintain a portal and keep it alive its content needs to be updated frequently notonly by information integration of different sources but also by additional inputs from humanexperts. The input view is defined byqueries to the schema, i.e. queries to the ontology itself.Similar to [Grosso, 1999]we support the knowledge acquisition task by generating forms out ofthe ontology. The forms capture data according to the ontology in a consistent way which arestored afterwards in the warehouse.Navigation view. To navigate and browse the warehouse we automatically generatenavigational structures by usingcombined queries for schema and content. First, we offerdifferent user views on the ontology by using different types of hierarchies (e.g.is-a, part-of )for the creation of top level navigational structures. Second, for each shown part of the ontologythe corresponding content in the warehouse is presented. Therefore especiallyusers that areunfamiliar with the portal are supported to explore the schema and corresponding content.(General) View. In the future we plan to explore techniques of handling updates on theseviews.

References[Angele, 2000] J. Angele, H.-P. Schnurr, S. Staab, and R. Studer. The times they are a-changin’ —the corporate history analyzer. InD. Mahling and U. Reimer, editors,Proceedings of the Third International Conference on Practical Aspects of Knowl-edgeManagement. Basel, Switzerland, October 30-31, 2000, 2000.http://www.research.swisslife.ch/pakm2000/.[Benjamins, 1999] V. R. Benjamins, D. Fensel, S. Decker, and A. G. Perez. (KA): Building ontologies for the internet.InternationalJournal of Human-Computer Studies (IJHCS), 51(1):687–712, 1999.[Decker, 1999] S. Decker, M. Erdmann, D. Fensel, and R. Studer. Ontobroker: Ontology based access to distributed and semi-structured information. In R. Meersman et al., editors,Database Semantics: Semantic Issues in Multimedia Systems, pages 351–369.Kluwer Academic Publisher, 1999.[Fensel, 2000] D. Fensel, J. Angele, S. Decker, M. Erdmann, H.-P. Schnurr, R. Studer, and A. Witt. Lessons learned from applyingAI to the web.International Journal of Cooperative Information Systems, 9(4):361–382, 2000.[Staab, 2000] S. Staab, J. Angele, S. Decker, M. Erdmann, A. Hotho, A. Maedche, H.-P. Schnurr, R. Studer, and Y. Sure. Semanticcommunity web portals. InWWW9 / Computer Networks (Special Issue: WWW9 - Proceedings of the 9th International World WideWeb Conference, Amsterdam, The Netherlands, May, 15-19, 2000), volume 33, pages 473–491. Elsevier, 2000.[Sure, 2000] Y. Sure, A. Maedche, and S. Staab. Leveraging corporate skill knowledge - From ProPer toOntoProper. In D. Mahling and U. Reimer, editors,Proceedings of the Third International Conference onPractical Aspects of Knowledge Management. Basel, Switzerland, October 30-31, 2000, 2000.http://www.research.swisslife.ch/pakm2000/.[Wiederhold, 1997] G. Wiederhold and M. Genesereth. The conceptual basis for mediation services.IEEE Expert, 12(5):38–47, Sep.-Oct. 1997.[Fernandez, 2000] M. F. Fernandez, D. Florescu, A. Y. Levy, and D. Suciu. Declarative specification of web sites with Strudel.VLDB Journal, 9(1):38–55, 2000.[Erdmann, 2001] M. Erdmann and R. Studer. How to structure and access XML documents with ontologies.Data and KnowledgeEngineering, 36(3):317–335, 2001.[Grosso, 1999] E. Grosso, H. Eriksson, R. W. Fergerson, S. W. Tu, and M. M. Musen. Knowledge modeling at the millennium: thedesign and evolution of PROTEGE-2000. InProceedings of the 12th International Workshop on Knowledge Acquisition, Modelingand Mangement (KAW-99), Banff, Canada, October 1999.[Stojanovic, 2002] L. Stojanovic, N. Stojanovic, and R. Volz. Migrating data-intensive web sites into the semantic web. InProceedings of the ACM Symposium on Applied Computing SAC-02, Madrid, 2002, 2002.


95/100

Annex 2

Creating a SEAL-based Web Site

The creation of a SEAL-based web site is a multi-step process. The genesis starts with thecreation of the ontology, which provides a conceptualization of the domain and is later used asthe content model of the portal.

Step 1 – Ontology design:Here, several tools come in handy, within KAON Ont-O-Mat SOEPprovides an editor with strong abilities regarding the evolution of the ontology. OntoEdit is acommercial tool that provides support for the full cycle of ontology development andadditionally allows to provide F-Logic axioms to refine the ontology.

Step 2 – Integrating Information: The next step towards the final web site is providing data.Here, we take a warehousing approach to amalgamate information coming from heterogeneousdata sources._

ÿ� RDF metadataUser-supplied HTML and PDF documents have to be annotated withmetadata based on the content ontology in order to be part of the SEAL portal. Thesedocuments can be located anywhere on the web and are made part of the portal usingKAON Syndicator, a component that gathers the meta data contained in resourceslocated on the web.

_

ÿ� Database ContentToday most large-scale web applications present content derivedfrom databases. KAON REVERSE is an application that provides visual means to mapthe logical schema of relational databases to the integrated conceptual model providedby the ontology [Stojanovic, 2002]. The user-supplied mappings are then used totransform the database content to ontology-based RDF.

_

ÿ� Peer-To-PeerAlso connectors to theEdutella peer-to-peer network, that provides anRDF-based meta-data infrastructure for peer-to-peer applications, are currentlyconstructed within KAON. SEAL portals can then be used to provide a web accessibleinterface to Edutella based Peer-To-Peer networks

Step 3 – Site design:We derive the previously mentioned navigation model and personalizationmodel from the ontology. Currently no extensive tool support for these tasks exist. Both modelsare derived from the ontology using F-Logic queries that are provided by the site administrator.

Navigation modelBeside the hierarchical, tree-based hyperlink structure which corresponds tothe hierarchical decomposition of the domain, the navigation module enables complex graph-based semantic hyperlinking, based on ontological relations between concepts (nodes) in thedomain. The conceptual approach to hyperlinking is based on the assumption that semanticallyrelevant hyperlinks from a web page correspond to conceptual relations, such asmemberOf orhasPart , or to attributes, likehasName. Thus, instances in the knowledge base may bepresented by automatically generating links to all related instances. For example, on personalweb pages there are, among others, hyperlinks to web pages that describe the correspondingresearch groups, secretary and professional activities(cf. figure A1.1).


96/100

Figure A1.1 - Templates generated from the web-site models

Step 4 – Web design:The derived models constructed in step 3 serve as input to the KAONPortal Maker, which renders the information in HTML. The implementation of KAON portalMaker adheres strictly to a model-view-controller design pattern. The ontology and the derivedmodels are encapsulated by an abstract data model and the presentation of the information iscreated using template technologies like JSP, ASP or XSLT. Default controllers are providedfor standard application logic like up-dating data and generating links to other presentationobjects. The reader may note that the default controllers can be replaced by custom-made con-trollers provided by the site administration. KAON Portals also provides default templates thatprovide the most often used representations for information objects (like list-entries, forms forweb-based data provision etc.) For instance, the AIFB portal includes an input template(cf.figure A1.1, upper part) generated from the concept definition ofperson (cf. figure A1.1,middle left) and a sheet like representation to produce the corresponding person web page(cf.figure A1.1, lower part). These default templates can easily be customized for special purposes.

References[Stojanovic, 2002] L. Stojanovic, N. Stojanovic, and R. Volz. Migrating data-intensive web sites into thesemantic web. InProceedings of the ACM Symposium on Applied Computing SAC-02, Madrid, 2002,2002.


97/100

Annex 3

Representative applications for corporate intranet andKnowledge Management

8.4 OntoBroker

OntoBroker is one of the first implemented tools powered by the use of ontology and Semantic Webtechnology. It contains three core elements: a query interface for formulating queries, an inference engineused to derive answers, and a webcrawler used to collect the required knowledge from the web. It targetsthe support for providing a service that can be used more generally for the purpose of knowledgemanagement and for integrating knowledge-based reasoning and semiformal representation of documents.The query formalism is oriented toward a frame-based representation of ontologies that defines the notionof instances, classes, attributes and values (D. Fensel, et al., 1998). OntoBroker has been successfullyused in several user case scenarios:

ÿ� Semantic Community Web Portals: A Portal for the Knowledge Acquisition Community

ÿ� Knowledge Annotation Initiative of the Knowledge Acquisition Community (KA)2

ÿ� ProPer: Human Resource Knowledge Management

URL: http://ontobroker.aifb.uni-karlsruhe.de/index_ob.html

8.5 OntoKnowledge

Content-driven Knowledge-Management through Evolving OntologiesEfficient knowledge management has been identified as key to maintaining the competitiveness oforganizations. Traditional knowledge management is now facing new problems triggered by the web, toname but a few, information overload, inefficient keyword searching, heterogeneous informationintegration, geographical-distributed intranet problem and so on. These problems will be tackled by themodern technology – called the Semantic Web Technology (Fensel, 2001 & to appear). On-To-Knowledge13 (OTK) project is an important player devoting itself to finding the content-drivenknowledge management solutions through evolving ontologies. It employs the power of the SemanticWeb Technology to facilitate knowledge management.

URL : http://www.ontoknowledge.org

Tool structureOn-To-Knowledge supports efficient and effective knowledge management by providing a toolenvironment powered by the Semantic Web Technology. It focuses onacquiring, maintaining, andaccessingweakly structured information sources:

ÿ� Acquiring: Text mining and extraction techniques are applied to extract semantic information fromtextual information (i.e., to acquire information). Tool support includes ontology extraction fromtext (OntoExtract and OntoWrapper)

ÿ� Maintaining: RDF, XML and DAML+OIL are used for describing syntax and semantics of semi-structured information sources. Tool support includes ontology editor (OntoEditor), and ontologystorage and retrieval (Sesame, RDF-Ferret), so as to enable automatic maintenance and viewdefinitions on these knowledge.

13 www.ontoknowledge.org


98/100

ÿ� Accessing: Push-services and agent technology support users in accessing the information. Toolsupport includes ontology-based information navigation and querying (RDF-Ferret), and ontology-based visualization of information (Spectacle).

In a nutshell, the complete layered tool environment of the OnToKnowledge is (figure A1.2): OntoExtractand OntoWrapper extract unstructured and structured textual information sources from specified domainsat the Internet or Intranet. The extracted information is pumped into the RDF-DB (Sesame), where it canbe edited with the OntoEdit tool. Finally, the RQL (RDF querying language) reasoning engine allows forthe querying of this database and delivers results to a user through a smart user- RDF-Ferret andvisualized by Spectacle.

Real-life ApplicationsThree real-life applications have been conduct during the course of the OTK project to fulfil tworequirements: - to identify the real-life requirement for the design of the tools and another way around - tosecure the usability of the tools for tackling the real-life problems.

Figure A2.1 - OnTo Knowledge The layered tool environment of the OnToKnowledge

BritishTelecom call center: Call center is the platform for companies to communicate with theircustomers and its market is growing 20% a year, with millions being spent on improving customerrelationships. Current call center technology lacks of the support of the operator in solving incomingrequests. The investment in call center technology can offer great rewards, including better customerservice, lower overheads, lower operational costs, and increased staff profitability. In the BT case study, asystem for supporting intranet-based virtual communities of practice is being developed, allowing theautomatic sharing of information. The system,OntoShare, allows the storage of best practice informationin an ontology and the automatic dissemination of new best practice information to relevant call centeragents. In addition, call center agents can browse or search the ontology to find information of mostrelevance to the problem they are dealing with at any given time. The ontology helps to orientate newagents and acts as a store for key leanings and best practices accumulated through experience. It providesa sharable structure for the knowledge base, and a common language for communication between callcenter agents.

ÿþýü

��ýüüý�

�ýþ��ý

��

�ý��ýüþ�� ü�� ü

�� ý��ý��þ��ü�

��ý��þ��ü��ý��ýü��

��üý��

�üý��þ��ü�

��

� � � � � �üý

� �� ü�� ü��ýü

� ��

��ý��ý


99/100

Swiss Life Applications: Two of the case studies were carried out by Swiss Life. One of it approachedthe problem of finding relevant information in a very large document about the International AccountingStandard (IAS) on the extranet. With the help of the ontology extraction tool OntoExtract an ontologywas automatically learned from the document, which significantly supports a user in reformulating aninitial query when it did not deliver the intended results. The second case study of Swiss Life is a skillmanagement application that uses manually constructed ontologies about skills, job functions, andeducation. This enables an employee to create in a simple way a personal home page on the company'sintranet that includes information about personal skills, job functions, and education. Using the ontologyallows a comparison of skills descriptions among employees, and makes it possible to automaticallyextend a query with more general, or more specialized, or semantically associated concepts.

Enersearch Applications: The case study done by EnerSearch AB focuses on validating the industrialvalue of the project results with respect to the needs of a virtual organization. The main difficulty with thecurrent web site is that it is rather hard to find information on certain topics because the current searchengine supports free text search rather than content-based information retrieval. To improve, the wholeweb site is annotated by concepts from an ontology which was developed using a semi-automaticextraction from the documents on the existing EnerSearch web site. The RDFFerret search engine is usedto extend the free text search to a search of the annotations as well. Alternatively, with the Spectacle toola user is able to get a search result arranged into a topic hierarchy which can then be browsed, allowingthe user to look for the required information in a more explorative way.

8.6 CoMMA (Corporate Memory Management through Agents).

Objectives

The COMMA projectaims at implementing a corporate memory management framework based on agenttechnology, that will capture, store and diffuse embedded knowledge of many types, in interactivesessions to employees.The CoMMA project will address particularly the following issues:Enhancing the insertion of new employees in the company by capturing experience and know-how fromelder employees. The facility and the rapidity to integrate new employees are a major issue in acompetitive market.Performing processes that detect, identify and interpret technology movements and interactions formatching technology evolutions with market opportunities to diffuse among employees innovative ideasrelated to technology monitoring activities.

COMMA project will design this corporate memory by merging most innovative technologies:

ÿ� Multi agent system,

ÿ� XML, RDF format,

ÿ� machine learning techniques

ÿ� COMMA project will provide new services, products and tools for knowledge management asdedicated agents able to achieve specific tasks for information retrieval

URL : http://www.si.fr.atosorigin.com/sophia/comma/Htm/HomePage.htm


100/100

Figure A2.2 - CoMMA Overall Architecture