W3C home > Mailing lists > Public > www-rdf-interest@w3.org > April 2002

Re: Ontologies, inference etc

From: Gerry Wolff <gerry@informatics.bangor.ac.uk>
Date: Tue, 23 Apr 2002 16:09:59 +0100
Message-ID: <008e01c1ead8$eff30040$37068f93@sees.bangor.ac.uk>
To: "SIG RDF" <www-rdf-interest@w3.org>
It has been suggested to me that the abstract of the article should be
posted. So here it is (at the end of this message).

Re "Does it have anything to do with RDF?", the ICMAUS framework is a
possible alternative to RDF and related approaches. The article contains
some comparisons between the two.




A computational framework called {\em information compression by multiple
alignment, unification and search} (ICMAUS) has been developed with the aim
of integrating concepts in computing, AI and cognitive science. This article
introduces the ICMAUS framework and describes how it may be applied in the
Semantic Web. At appropriate points throughout the article, comparisons are
made with existing proposals for the Semantic Web.

In general terms, the main benefits of these ideas are a simple, versatile
format for the representation and integration of diverse kinds of knowledge
and a unified framework for processes of pattern recognition, information
retrieval, probabilistic reasoning and others.

The framework and its implementations support the representation and
integration of class-inclusion hierarchies with part-whole hierarchies in a
simple, intuitive manner. They also support the representation and
integration of other forms of knowledge, including context-free grammars,
context-sensitive grammars, discrimination nets, tries, if-then rules, and

From these kinds of structure, inferences may be drawn, each with an
associated probability. Kinds of probabilistic reasoning supported by the
framework include probabilistic `deduction', abduction, chains of reasoning,
nonmonotonic reasoning and `explaining away'.

At the heart of the framework is an improved version of `dynamic
programming' for finding full matches and good partial matches between
patterns.  This enables the framework to deliver useful results with
information that contains errors or is incomplete.

With some further development, potential applications include planning and
problem solving, understanding and production of natural language,
abstraction of knowledge from raw data and the unsupervised learning and
organisation of knowledge.

Throughout the article, examples are illustrated with output from the SP61
computer model.
Received on Tuesday, 23 April 2002 11:10:39 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 7 January 2015 15:07:40 UTC