Intelligent Critiquing of Specifications based on Ontologies
All formal modelling uses a particular mathematical language to describe
a chosen domain. In doing this we can make mistakes related to the
mathematical language (if, for example, we write a non-terminating recursion
using a logic programming language) or we can make mistakes in describing the
domain (for instance, we can define an ecological model in which animals
photosynthesise). The latter type of mistake is difficult to detect because
it requires subjective knowledge about correct forms of domain description to
be applied to the model description. We call this sort of mistake a
conceptual error.
In the past, it was difficult to find explicit guidelines for domain
description. In the recent years ontologies have become popular in the
KBS community as a way of promoting knowledge sharing and reuse.
Some of these provide formal constraints on the way target domains should
be described, via axioms which restrict the interpretations that the
ontology's constructs could have.
We are investigating how these can be used to detect
conceptual errors in specifications that are based on these ontologies.
