WebKB refers to the knowledge base (KB) servers WebKB-1 and WebKB-2.
KB servers are not Web search engines; they are Web-accessible
knowledge base
(management) systems (KBSs / KBMSs).
KBMSs are database
management systems (DBMSs) that, unlike relational DBMSs, object-oriented DBMSs
and deductive DBMSs, permit end-users (e.g., Web users) to dynamically modify or add
a large number of conceptual schemas/definitions in their KBs and hence do not limit
end-users to predefined kinds of data.
KBMSs can support logic-based semantic networks and use logical inferences to
check knowledge entering and help knowledge retrieval. WebKB-2 is a large-scale KBMS
that has KB sharing protocols, special notations, and a large default
ontology (i.e., a large set of default conceptual schemas/definitions), to ease
knowledge entering and sharing.
WebKB can also be seen as an on-line private/shared knowledge-based annotation tool. The expression "knowledge-based annotation tool" means that WebKB is a knowledge representation tool extended to permit the indexation of information (any part of any Web-accessible document) with knowledge statements and then, the retrieval of this information via knowledge queries. The distinction "private/shared" means that both the distributed and centralized approaches for knowledge construction, sharability and reuse have been considered.
Though this tool may be used for manually indexing and then retrieving documents, it was meant for retrieving precise information, not simply documents. Hence, it is unrelated to classic document retrieval systems (e.g. Web Search engines such as Altavista or Yahoo) and to the Web->KB project which aims to automatically extract some knowledge from Web documents for classifying them and then easing their retrieval and browsing.
WebKB-1 permits Web users with
some aptitude in knowledge representation to store, organize and retrieve
knowledge or
document elements (DEs)
within Web-accessible files.
Knowledge may be asserted via various notations
(Conceptual Graphs and
simpler notations we designed),
mixed with other elements with a document (e.g. text or HTML elements),
used for indexing any part of any Web-accessible document (via
an indexation language), and
loaded, queried or combined via a
script language.
Thus, this tool permits knowledge engineers to build and exploit
knowledge bases and their links to source/generated documents. WebKB-1 may
also be used as a private precision-oriented annotation/retrieval tool.
In this approach, apart from the fact the knowledge is
Web-accessible, no support is provided for its reuse and sharability.
The users must know the URLs of documents countaining knowledge and send these
URLs to WebKB before making requests on that knowledge. However, some basic
text search and Web-exploration commands are provided.
The Word Wide Web Consortium (W3C) now
proposes the Resource description Framework
(a simple semantic network model) and its associated XML notation (RDF/XML)
as a recommendation for publishing knowledge within Web-accessible documents.
Though current RDF tools are still simple, some
will ultimately have the functionnalities of WebKB. However, RDF has
problems that complexify its use and limits its reuse:
1) its expressivity is poor (even with OWL),
2) its XML-based notation is difficult to read and cumbersome to write,
3) XML, and therefore RDF/XML, does not permit RDF statements to be mixed with
other textual/XML elements within a document.
Though we have proposed conventions and an
ontology to extend the expressivity of RDF and better allow its reuse,
it will take many years for similar propositions to be considered by the W3C.
The problem is the same for the last third point and the situation is definitively
hopeless for the second point. Interfaces or higher-level notations (such as
those of WebKB) are and will be necessary to build and update large or complex
RDF knowledge bases. When RDF expressivity
rises sufficiently for a decent use and reuse, translators from/to RDF will be
added to WebKB. By that time, RDF statements on the Web will be collected by
Web search engines, which will greatly improve knowledge sharability
in this distributed approach.
WebKB-2 permits Web users to store, organize and retrieve knowledge in a large single knowledge base on WebKB server machine. The above cited knowledge representation notations can be used here too but updates within loops are forbidden (because the whole knowledge base currently has to be locked while an update is made). Simple interfaces exploiting the content of the knowledge base are provided to permit easy accesses and updates to the knowledge base. Updates are regulated by permissions and protocols to ensure harmonious collaboration between the users. Click here to access an article (25 pages) on this point.
3. Bibliography
Published articles and documentation
Conventions and ontologies. Proposal of conventions and ontologies for RDF users.
Comparison of WebKB with different kinds of tools.
Comparison of WebKB with some other CG tools.