Domain Ontology Learning for Question Answering System in Network Education

Abstract

The BITQAP question answering platform is ontology-based semantic question answering system. In ldquoComputer Organization and Architecturerdquo course, we apply the platform as CAI tools to answer questions. In the paper, we proposed methods to extract the concepts and relations from course corpus. Term extraction is the first step of ontology learning. The Chinese domain terms have three features: Coupling, Domain Relevancy and Domain Consensus. In the paper, these features be modeled and integrated to evaluate the terms. In concept learning process, the CBC (Clustering by Committee) is adopted to learn concepts. In relation learning process, snow ball technique to be used to acquire hyponymy relation, and HowNet-based method to be used to extract general relation. Experimental results show that the methods of ontology learning in the paper is efficient and is helpful to solving of semantic problems in question answering system.