Concept-Base Construction for a Word Association-System: A Method of Deciding Word Attribute Weights by Logical Relations and Attribute Reliability

摘要

We think that in order to realize computers to understand natural language, an association-system which outputs words related to input words is needed. This study aims to construct a concept-base which is a main element of the association-system. The association-system consists of the concept-base automatically made from several electronic dictionaries and an algorithm measuring the degree of association between words. In the concept-base, a word is defined by a set of words (attributes) expressing the feature of the meaning of the word, with their weights decided by their frequencies in the dictionaries. There are, however, many unsuitable attributes in the concept-base, and their weights are unreliable because of the automatic construction. In this study, the automatically constructed concept-base is the subject of refinement. This paper proposes the method of deciding the weights by logical relations, and the attribute reliability as a refining method. Moreover, this paper shows effects of the proposed method by evaluation with human sense and with the degree of association.