Message 9016 of the SUO list

Subject: SUO: CG representations for WordNet
Date: Wed, 15 Jan 2003 09:34:33 -0800
From: but kindly forwarded by Adam Pease
In reply to: Richard Cooper's question

Rich, Adam,

 > ... If word senses (types) are in a generalization lattice,
>  does that mean that every node in the lattice has one or more CGs?
>  ... is there a database of CGs that corresponds to the
>  WordNet entries?

Many primitive relations (not just the very few listed in WordNet)
can be associated to categories for situations (states or processes),
e.g. case/spatial/temporal relations, and most instances of situation
categories can be source/destination of most of these relations.
The following URL permits to browse an ontology for such relations:

Schemas may be associated to categories to represent the most common
basic relations associated to these categories. The knowledge server
WebKB-2 ( exploits schemas to generate cascasding menus in order
to help users create normalized knowledge representations.
Such schemas do not need to be associated to all categories, only relatively
high-level categories (these schemas are then inherited by more specialized
categories). FCG representations of such schemas are at:

The default ontology of WebKB-2 (the one accessible at and used
in the above cited examples) is an integration of several top-level ontologies
with a correction and extension of WordNet 1.7. For details, see
Each category identifier is prefixed by its source, e.g. Sowa, although
the prefix "wn" is left implicit for WordNet categories.
Many options to browse this ontology are provided by

It is also available for download in the FO format, and in slightly
impoverished ways, in DAML/RDF, CGIF and WordNet.
This ontology was announced on the CG and SUO lists on the 27/06/2002

I should note that this ontology only integrates the noun-related part of
WordNet 1.7 since WebKB-2 is designed for normalized and explicit
knowledge representation by people, not for natural language parsing.
In WordNet, most categories for actions are both in the verb part and
the noun part (try "cut" or "eating" for example). The generalisation
hierarchy for the noun part is more adequate for knowledge representation.
WordNet 2.0 is expected to provide the (identity/synonymy?) links between
the categories in the noun part and those in the verb part. If this is the
case, both parts will be mergeable into a single generalization hierarchy.


Dr. Philippe Martin
Senior Research Fellow at the Distributed Systems Technology Center
                              (DSTC Pty Ltd;  DSTC is W3C's Australian Office)
Address: Griffith Uni, School of I.T., PMB 50 GCMC, QLD 9726 Australia
Email:;  Fax: +61 7 5552 8066