Overview of WebKB tools and examples

Dr. Philippe A. MARTIN

Abstract.   The WebKB set of tools - or simply WebKB - allows you to store, organize and retrieve knowledge or document elements (DEs) in Web-accessible files.
This document presents the tools and example documents proposed by the WebKB interface.


Fig. 1.  General interactions between WebKB tools

There are a number of tools that can be initiated from the main page of WebKB.

The "classic information retrieval/handling tool" allows you to apply and combine Unix-like text processing commands - such as cat, grep, awk and diff - on Web-accessible files of your choice. To use this tool, just select, paste or type commands or a command example, and submit.
Commands are sent to the WebKB server which executes them and returns an HTML document containing the commands and their results, or their results only.
Files containing commands (possibly mixed with other DEs) may be "loaded" in the WebKB processor, via a call to a command "load", "run" or "call" (these loading commands lead to different kinds of query/answer visualisation and, if the file is a script or a CGI server, have different ways of sending parameters to it).

The "knowledge-based information retrieval/handling tool" allows you to apply and combine knowledge building/retrieving commands.
Three "query context" fields provide a way to globally specify what kind of of knowledge and indexed elements should be searched, and how the results should be presented. Such specifications can also be done via commands (in the "query" area or in files).
Commands are executed in the order given by the user in the query area or in the loaded files, therefore at least some knowledge building/loading commands must be specified before knowledge retrieving commands (otherwise, they won't retrieve anything !). Similarly, types must be declared before they are used in knowledge representations or queries.
This tool also lists Unix-like text processing commands in its command menu since it will come in handy to combine theses with knowledge processing commands.

The "tool to index DEs by knowledge representations" generates the notation for an indexation of a DE by a knowledge representation.
You must specify (i) the URL of the document containing the DE and (ii) the DE itself, i.e. its content or a reference to it, and the way it is referenced or described: via a "URL", a "section title" if the DE is a section, no description if the DE is the "whole document", an "HTML description" or a "text content" for other kinds of DEs. If necessary, the occurrence of the DE in the document may also be provided.
You must then specify the knowledge representation or the URL of a file which contains it (and only it).
It is handy to copy-paste the DE from its document into the query area, and to copy-paste the generated notation into a Web-accessible file which will later be used via a "load" command. An example of such a file is given by the interview indexation.

Similarly, the "tool to connect DEs by conceptual relations" generates the notation for a connection of two DEs by a conceptual relation. A list of relation types is proposed but you may use any other that you have specified in an ontology. A connection example may be found in the interview indexation.

The "conceptual graph textual editor" hastens the construction of a CG which may then be copy-pasted into a Web-accessible file or in the query area of the "knowledge-based retrieval/handling tool".
First add a concept, then a relation, and then a relation to the concept you like by specifying it via the variable in its referent. Types for concepts or relation may be searched in a given ontology via the "hierarchy browser" (see below).
The CG may contain HTML marks and it may be graphically visualised.

Example documents

At present, knowledge can only be stored in documents.

The top-level ontology proposed by WebKB includes about 200 concept types and 200 relation types from various other top-level ontologies. Since it defines consistency constraints and provides lots of general purpose relations, it is handy when defining an ontology to begin by a command loading this top-level ontology, and then to specialise it. WebKB provides two examples of application ontologies, one about KADS and the other about Road accident analysis.
Such application ontologies may then be used via a loading command in files representing facts about these applications. Such files are illustrated by the following examples.

Finally, the script scr.html illustrates different kinds of knowledge representations, queries, generations and combinations, while the script whatIs.html shows how parameters may be used in scripts.