关键词: 叙词表自动构建, 词族, 概念语义网络, 社会网络分析, 层次聚类方法, 词素后方一致

Abstract:

[Purpose/significance] This article proposes a new method to recognize word clusters which could be used for semantic tools construction and query expansion. This method can reduce the load of experts' cognition burden and promote the efficiency of generating and updating semantic tools. [Method/process] This paper used Island algorithm in the social network analysis to discover word clusters in a sematic network of concepts which was generated through the word co-occurrence analysis and sematic similarity computing. By taking the finance field as an example, this article compared the proposed method with the hierarchical clustering method and the "same morpheme" method. [Result/conclusion] It is discovered that Island algorithm is better than hierarchical clustering algorithms in word cluster recognition. Island algorithm and the "same morpheme" method have their own advantages, so they can be used in combination and complement each other.

Key words: automatic thesaurus construction, word clusters, semantic network of concepts, social network analysis, hierarchical clustering, the "same morpheme", method