Semi-automatic Integration of Knowledge Sources

Abstract

Integration of knowledge from multiple independent sources presents problems due to their semantic heterogeneity. Careful handling of semantics is important for reliable interaction with autonomous sources. This paper highlights some of the issues involved in automating the process of selective integration and details the techniques to deal with them. The approach taken is semi-automatic in nature focusing on identifying the articulation over two ontologies, i.e., the terms where linkage occurs among the sources. A semantic knowledge articulation tool (SKAT) based on simple lexical and structural matching works well in our experiments and semi-automatically detects the intersection of two web sources. An expert can initially provide both positive and negative matching rules on the basis of which the articulation is to be determined and then override the automatically generated articulation before it is finalized. The articulation may be stored or generated on demand and is used to answer customer queries efficiently.