HCHIRSIMEX: An extended method for domain ontology learning based on conditional mutual information

Abstract

This paper presents HCHIRSIMEX, an extended version of our previous algorithm HCHIRSIM for building domain ontology from web corpus. The new version introduces a novel measure based on the Conditional Mutual Information (CMI) statistic method to define the taxonomic relations and the similarity between selected concepts. By using this method, the ontology extracted by HCHIRSIMEX is more concise and contains a richer concept knowledge base compared with the previous version HCHIRSIM. To evaluate our new algorithm effectiveness, we apply the two algorithms and Sanchez et al. algorithm in Finance domain ontology constructed from the web. Then, we compare the obtained concepts with those on the “Financial glossary” provided by Yahoo.com.