[ieee 2010 8th ieee international conference on pervasive computing and communications workshops...
TRANSCRIPT
SPARK: A Service Personalisation Architecture
Gilbert Cassar
Centre for Communication Systems Research University of Surrey, UK
Email: [email protected]
Abstract-The current model of the Service Oriented Architecture (SOA) does not directly support possible scenarios for pervasive environments where services can be offered to/from the smart personal devices carried around by different users. Service description technologies such as WSDL provide descriptions of what a service is and what it can do, however with these kinds of descriptions machines still cannot easily interpret different concepts described through technical attributes. If more expressive technologies are used to describe a service, machines become able to identify different components describing a service. In pervasive environments users move around with Personal Agents operating on their portable devices and as mobile devices become more powerful, users are becoming able to host services on their mobile devices too. This project proposes a Service Personalization Architecture that will enable machines to understand the functionality provided by different services and use this knowledge together with the attributes provided by the users' smart devices in order to provide service recommendation, service composition, and service personalisation to the users.
I. INTRODUCTION
Although the current model of the Service Oriented Architecture (SOA) has served its purpose well in the past years,
the future of software-based services is moving towards more pervasive and personalised environments where each client is targeted as an individual with personal preferences,
lifestyle, agenda and goals. In order to embrace this new approach and to exploit the available information to provide users with more personalized functionality, the present
architecture needs to be changed. The future of pervasive computing is envisioned to
provide every person with an Intelligent Personal Mobile
Lifestyle Assistant which may take many forms [2]. The
problem with this approach is to make the Personal Agent
ubiquitous as the user migrates from one device to the other [3]. In order to support this interaction between users and devices, a rich distributed service framework is needed in order to describe different aspects for personalisation [2]. The services will use the client's personal information (de
rived from user profiles, context-data, device profile, and network attributes) to personalise services and adapt content
to the user's needs. The problem we identify in this project is that the current
technologies used in the SOA are not expressive enough and
are limiting the quality of user experience when discovering, selecting, and consuming services. Service descriptions
978-1-4244-5328-3/10/$26.00 ©2010 IEEE 857
are not usually defined with machine-interpretable, non
technical and application-domain metadata. If concepts such as the type of data, goal, adaptability, and availability of
services are defined in a machine-interpretable manner, a search and discovery mechanism can provide more than
just a simple match. In order to use machine-interpretable
service descriptions, semantic web technologies could be used in order to create a common machine-interpretable
way for describing a service and its components and also a declarative structure for queries.
II. SERVICE PERSONALISATION VS CONTENT
PERSONALISATION
Personalisation is an approach which accommodates the differences between individuals. Rather than having a stan
dard service or content presentation which is accessed by
all users, we look at a service as a set of functionalities and content as a set of data and we choose which functionalities
and data are most suitable for the user. "Personalisation is about building customer loyalty by developing a meaningful
one-to-one relationship with clients, by understanding the
needs of each individual and helping satisfy a goal that efficiently and knowledgeably addresses each individual's needs in a given context" [4].
A. Content Personalisation
Content Personalisation is greatly associated with the technology of adaptive presentation generation (e.g. adaptive hypermedia presentation generation). Content Personalisa
tion aims to present the users with the data that is most
relevant to their own interests, experience, and profile. When
a huge amount of data is available, the user can easily end up roaming through links and pages until the relevant data is found. Content Personalisation relies on the user's profile
to determine what the user's needs are and what level of information would be suitable for the user. Then it presents the user with the appropriate content and presentation (data
can be presented in a manner deemed appealing for the user according to the profile) adapted to the user's needs.
B. Service Personalisation
Service Personalisation deals with the functionality and
the accessibility of a service. Service Personalisation starts
when a search and discovery mechanism starts looking for
the best matching service(s) that can accomplish that user's goal. A service will always have a core functionality, however it can also define some side functionalities which are
optional and adaptable for each user. Service Personalisation selects and assembles together different functionalities in order to tailor services for different users. In more complex
scenarios, different services can be composed together by a composition engine in order to present the user with an
automatically generated service [1].
III. SERVICE PERSONALISATION ARCHITECTURE
In this project we propose a service personalisation architecture that will enable machines to interpret and reason
the functionality provided by different services and use this knowledge to provide Service Recommendation and Personalisation functionalities. Figure I shows the core components in the architecture. The architecture uses ontology based service descriptions and a reasoner to search for relevant
services. Probablistic semantic approaches will then be used to recommend services to the user and machine learning
techniques such as Bayesian Networks are used to provide
the composition and personalisation of services based on the user's profile which is kept up-to-date by the Personal
Agent of the user. The architecture consists of the following components:
• Service Description Ontology
• Service Repository • Service Personalisation and Composition Engine
• Personal Agents
• Content Personalisation
The Service Repository will then allow service providers
to publish their services in the repository, as long as they abide by the rules of the Service Description Ontology (i.e.
an extended implementation of a UDDI with semantic search
and discovery mechanism). The Service Personalisation and Composition Engine (SPACE) interacts with the user's Per
sonal Agent and with the Service Repository and uses in
formation from both entities to create personalised services. When a search is initiated, SPACE requests information from
the user's profile (stored on the Personal Agent of the user)
in order to facilitate personalisation. SPACE uses a reasoner
and filtering techniques in order to provide service recommendation, service orchestration/choreography, and service personalisation. The Personal Agent is a software running on
the device of every user on the network and is used to interact with SPACE automatically. Different data mining and clustering techniques are considered in the Personal Agent to create a user profile while the user is searching, browsing, and accessing services on the device [4]. After Service
Personalisation, the Personal Agent accesses the Service and exchanges user profile information with the service to adapt the content to the user's preferences. Content Personalisation
is very closely related to service personalisation. This stage
858
Figure 1. Service Personalisation Architecture
will not be explored in detail in this project but it is the step that follows after a service is personalised.
IV. CONCLUSIONS AND FUTURE WORK
The Service Personalisation Architecture proposed in this
paper facilitates the integration of mobile services into SOA and also provides an enhanced framework on which to build mechanisms to provide service recommendation, service composition, and service personalisation. Future work will
concentrate on designing efficient algorithms for service search, discovery, recommendation and composition in the
context of personalised mobile services.
ACKNOWLEDGMENT
This PhD project is supervised by Prof. Klaus Moessner and Dr. Payam Barnaghi at the Centre for Communication Systems Research, the University of Surrey.
REFERENCES
[1] Adam Barker, Christopher D. Walton, and David Robertson. Choreographing web services. IEEE Transactions on Services Computing, 2(2):152-166, 2009.
[2] Patricia Charlton and Myriam Ribiere. Rich service description for a smarter lifestyle. In AAMAS '03: Proceedings of the second international joint conference on Autonomous agents and multiagent systems, pages 512-519, New York, NY, USA, 2003. ACM.
[3] Yuhong Feng, Jiannong Cao, I. Lau, and Xuan Liu. A selfconfiguring personal agent platform for pervasive computing. In Embedded and Ubiquitous Computing, 2008. EUC '08.
IEEEIIFIP International Conference on, volume 1, pages 438-444, Dec. 2008.
[4] E. Frias-martinez, S.Y. Chen, and Xiaohui Liu. Survey of data mining approaches to user modeling for adaptive hypermedia. Systems, Man, and Cybernetics, Part C: Applications and Reviews, IEEE Transactions on, 36(6):734-749, Nov. 2006.