helping people find content … preparing content to be found enabling the semantic web joseph busch
TRANSCRIPT
Users cannot type in the complex queries needed to find all the relevant items... But this can be done automatically.
Search Failure
19% Character errors. (Young, et al)
40% Vocabulary errors. (Seaman)
20% Index confusion.
21% Successful (Nielsen)
40%20%
19%21%
Search Solution
Generate more consistent content to search on.
Correct user errors.
Map the language of users to the language of the target content.
Search Alternatives
Personalization Content needs to be tagged with attributes that map to user categories
Analytics Users don’t follow predictable & consistent pathways
Taxonomies Automatically generated taxonomies reflect ambiguities of natural language
Syndication Requires subscriber profiles, well-categorized content, & managed rules
Solution for Search Alternatives
Predictable standardized structures, and
Consistent semantics to work on
… so machines can understand it.
Berners-Lee’s Semantic Web
Formatting content so that machines can understand it.
Use XML/RDF: Infinitely flexible markup language.
Process content in many more ways than simply for viewing it.
Problem: Mostly syntax … not semantics (in the human sense of meaning, i.e., language)
XML is a Grail-like Object
XML is just a means for encoding information—an envelope standard. The real value is still in the information that you put in the envelope.
Filling XML placeholders such as <meta>, <subject>, and <maker> requires semantic information management.
Soergel’s SemWeb Proposal
System of integrated access to data on concepts and terminology.
Bring together variety of sources that exist largely in separate worlds, including dictionaries, thesauri, classification schemes, etc.
Federated system with multiple collaborators.
Common interface to all concept & terminology knowledge bases on the Internet.
The Real Semantic Web
Namespace for uniquely identifying a semantic scheme & each concept within each scheme.
Broad template or conceptual schema for holding all types of semantic information & specifying relationships among them.
Definitions of services for interacting with the System.
Vocabulary Markup Language (VocML)
XML schema for the Semantic Web.
Broad template for structured representation of semantic schemes. Dublin Core metadata.
Tags and syntax for uniquely identifying each concept.
Typed relationships (hierarchical, associative, etc.)
Typed notes.
Networked Knowledge Organization Systems nkos.slis.kent.edu
<?xml version="1.0"?>
<!DOCTYPE VocML SYSTEM "VocML.dtd“>
<VocML version=”1.1“>
<SrcVocab>
<SVHeader>
<dc:Title>DFSIC-1998</dc:Title>
<dc:Source>Standard Industrial Classification (1987)</dc:Source>
<dc:Creator>Interwoven</dc:Creator>
<dc:Contributor>U.S. Department of Commerce</dc:Contributor>
…
<workNum UIDprefix=”DFSIC-1998” DisplayTitle=”Standard Industrial Classification” BriefDisplay=”SIC”>
</SVHeader>
<SVTerm UID=”DFSIC-1998::0139” CCID”104:43”>
<label>Field Crops, except Cash Grains, not elsewhere classified</label>
<definition>Establishments primarily engaged in the production of field crops, except cash grains, not elsewhere classified. This industry also includes establishments deriving 50 percent or more of their total value of sales of agricultural products from field crops, except cash grains (Industry Group 013), but less than 50 percent from products of any single industry.</definition>
<cla>0139</cla>
<typedRelation UREF=”DFSIC-1998::013” UTYPE=”Z39.19-1980::2" Name=”BT”>
<typedRelation UREF=”DFSIC-1998::013900” UTYPE=”Z39.19-1980::3" Name=”NT”>
…
Dublin Core
Unique ID
Typed Relationships
The Holy Grail is ...
Accurate information automatically processed so that it can easily be found and used for applications.
A rich web of linked information, with markup allowing machines to route relevant information to the audiences that value it most.
Metatagging
The hard work is mining content to extract key information: Recognize the mentions of people, organizations, places,
and things.
Infer the subject matter.
And putting it into formats with standard labels for effective exploitation.
Raw Content
• unstructured text
• untagged data
Semantic Content Management
Relevant Information
• found items
• granular text
User Queries
• database search
• text search
Structured Content
• metadata
• XML/RDF
Tag It
Exploit It
Vocabularies
Exploiting the Semantic Web
Route content to audience segments that value it most.
Link mentions of people, organizations, places, and things to other information related to those entities.
Populate portal directories.
Precisely search heterogeneous content items.
Predictions
VocabularyML. Semantic standard for unique identifiers (a namespace) for
people, organizations, places, and things and the relationships among them.
See: nkos.slis.kent.edu
Technologies that enable the persistent naming of the information inside XML envelopes.
Generation of enormous value through interoperability among web applications.
Joseph A. Busch Content Intelligence Evangelist
ASIST President, 2001
415-778-3129fax 415-778-3131
Moving business to the Webwww.interwoven.com