Template Based Attribute Value Words Acquisition in Entity Attribute Knowledge Base Construction

Abstract

In the task of entity attribute knowledge base construction, each attribute has many candidate value words. It is impossible to get these candidate value words manually. This paper proposed a template based attribute value words acquisition method, which could acquire the candidate value words of attribute from real corpus automatically. The acquisition of candidate value words were driven by attribute. The candidate value words could be obtained in the context which described the attribute. The expression of a particular text generally had a fixed form. The form could be extracted as a template for the acquisition of candidate value words of attribute. This paper gave the definition of the template and proposed an automatic template generating method. Based on the templates, this paper proposed a procedure of attribute value words acquisition which was constrained by the known attribute words. It extracted attribute value words from annotated corpus by the method of the template instantiation. According to the relationship between attribute word and attribute value word in the template, the relation-net of attribute words and attribute value words could be organized directly. The experimental result showed that the proposed method performance well on the real corpus.