Automatic Construction of Domain Ontology: A Hybrid Method Based on Chinese Encyclopedia Resource ⋆

Domain ontology plays an increasingly important role in a variety of fields, such as information retrieval, artificial intelligence, etc. However, its construction still remains a considerable challenge. In this paper, how to automatically create domain ontology is discussed and a hybrid method based on a famous online Chinese encyclopedia resource (Hudong.com) is proposed. We define domain ontology as domain concepts with four relationships: KindOf relationship, InstanceOf relationship, AttributeOf relationship and SynonymsOf relationship. Taking the advantage of hierarchy structure in Hudong.com, concepts of a domain ontology and KindOf relationship are firstly extracted. An entropy-based method is proposed to gain domain concepts. Then the rest relationships are obtained based on these concepts by using three methods respectively. Furthermore, an ontology of popular science domain is built using the proposed method, which contains 13,479 classes, 598,062 instances and 344 attributes. The experimental results show that good extraction accuracy is obtained without much manpower. Besides, the built domain ontology outperforms some existing ones, such as NASC’ plant ontology with respect to the construction cost and the richness of structured information.