WebKB (http://www.webkb.org) is a knowledge base server that permits
internet/intranet Web users to add, complement or annotate knowledge into
a shared knowledge base. Such a tool is useful when information of various 
kinds need to be interlinked or compared to each other, or retrieved 
via a question-answering system making inferences. This is the case for 
some corporate memories, structured catalogs (e.g. Yellow-Page-like 
catalogs or those in auction/advertisement Web sites), or more generally
for "knowledge" repositories. 

As opposed to relational or object oriented databases, information in 
knowledge bases (KBs) can be freely and dynamically structured by
end-users. The drawback is that knowledge is difficult to enter and
automatically compare or share. WebKB seems to be the first KB server
significantly reducing this difficulty. To do so, we have developped
and implemented (i) expressive and high-level input/output formats,
(ii) easy-to-use HTML-based interfaces, 
(iii) knowledge sharing protocols,
(iv) a merge, correction and extension of several top-level ontologies
and the natural language ontology WordNet, 
(v) the dynamic generation and combination of forms-to-fill based on 
schemas associated to categories (and we have written these schemas).
During the last two months, we have particularly focused on this 
last point, improved some parts of the code and documentation, and
developped the structure for a knowledge repository on IT resources
(see http://www.webkb.org/kb/it/d_IS.html). We have also integrated the 
DOLCE ontology (which is one of the candidate top-level ontologies
for the Semantic Web) but realized it is too specific to be really
usable by end-users and for restructuring a natural language
ontology such as WordNet. We have begun an article on our 
transformation of WordNet into a genuine lexical ontology usable
by knowledge-based applications, and we will submit it to 
WWW 2003 on November 15.

The implementation of WebKB-2 began in January 2000 and hence
its IP fully belongs to DSTC. WebKB-2 is now nearly mature for
commercial applications and, in our opinions, the most actually
usable "genuine" knowledge representation tool. We stress the
word "genuine" to differentiate our approach (where the meaning of
knowledge statements has to be explicit and hence can be exploited
for inferencing purposes) from semi-structured semantic network 
approaches (e.g. "concept maps") where free text can be used,
very few semantic checks are made, and hence entering information
is easier but much less inferencing can be done.
We have strived to developp a system permitting end-users to
enter knowledge precisely/explicitly and benefit from it. Now,
we may try to incorporate a less structured approach and see
if this is worthwhile. This would permit to incorporate 
current RDF resources on the Web without having to manually
correct them and fully integrate them to the existing KB
but the benefit of this next step is yet unclear.
The knowledge matching, retrieval and presentation mechanisms 
will also have to be extended.

An immediate next step is to prepare a specific application to
convince one or more Australian auction/advertising sites 
(e.g. sold.com.au or www.yellowpages.com.au) to incorporate
an adapted version of WebKB-2 to their Web sites, as a 
complement approach to their database-based approaches.