Detection of Affectively Comparable Term Using Hierarchical Knowledge and Blog Snippets

抄録

This paper describes a method for detecting affectively comparable terms. Comparable terms are often handled as sample objects instance in order to enrich linguistic expression, and using such terms explains and describes descriptions well. Coordinate terms in hierarchical knowledge are potentially comparable terms. Hierarchical coordinate terms are however sometimes affectively inappropriate as comparable term, because hierarchical knowledge is constructed by using only semantics without affections. We obtained the affections of terms obtained from blog and innovated them into hierarchical knowledge in order to detect affectively comparable terms. We conduct experiments to detect affectively comparable terms and discuss results, from which, we confirmed that affectively comparable terms could be detected by our proposed method. We deem detected affectively comparable terms to be applicable to creating artificial intelligence realizing intuitive interaction.