An inductive approach to assertional Mining of Web Ontology Revision.
Chieko Nakabasami · Rules and Rule Markup Languages for the Semantic Web · 2002
This paper proposes an inductive learning method for maintaining a web-based ontology by incorporating newly generated concepts from assertional knowledge (A-Box). The ontology used in this approach is represented by DAML+OIL. This ontology is translated into a form acceptable for the FACT system, a Description Logic (DL) reasoner, and is compiled into a knowledge base as a T-Box, a terminological knowledge description. Inductive learning is used for integrating the A-Box, where positive and negative examples submitted by human users are stored. Inductive Logic Programming (ILP) is used in order to induce concepts consistent with positive examples and to exclude negative ones. Such induced concepts are explored in order to find where they are positioned in the concept hierarchy in the T-Box, and the original ontology is revised. ILP can provide new concepts for DLs even though they may have richer expressiveness since DL is a decidable fragment of first-order logic. The induced concepts could be also utilized for predicting novel assertions from human users.