Hierarchical Structure in Semantic Networks of Japanese Word Associations

Hierarchical Structure in Semantic Networks of Japanese Word Associations * Maki Miyake a, Terry Joyce b, Jaeyoung Jung c, and Hiroyuki Akama c aOsaka University, 1-8 Machikaneyama-cho, Toyonaka-shi, Osaka, 560-0043, Japan bTama University, 802 Engyo, Fujisawa-shi, Kanagawa-ken, 252-0805, Japan cTokyo Institute of Technology, O-okayama, Meguro-ku, Tokyo, 152-8552, Japan mmiyake@lang.osaka-u.ac.jp terry@tama.ac.jp {catherina, akama}@dp.hum.titech.ac.jp Abstract. This paper reports on the application of network analysis approaches to investigate the characteristics of graph representations of Japanese word associations. Two semantic networks are constructed from two separate Japanese word association databases. The basic statistical features of the networks indicate that they have scale-free and small- world properties and that they exhibit hierarchical organization. A graph clustering method is also applied to the networks with the objective of generating hierarchical structures within the semantic networks. The method is shown to be an efficient tool for analyzing large-scale structures within corpora. As a utilization of the network clustering results, we briefly introduce two web-based applications: the first is a search system that highlights various possible relations between words according to association type, while the second is to present the hierarchical architecture of a semantic network. The systems realize dynamic representations of network structures based on the relationships between words and concepts. Keywords: Network analysis, Graph clustering, Japanese word associations. 1. Introduction As an approach to deepening our understanding of lexical knowledge, many areas of cognitive science, including psychology and computational linguistics, are seeking to unravel the rich networks of associations that connect words together. Key methodologies for that enterprise are the techniques of graph representation and their analysis that allow us to discern the patterns of connectivity within large-scale resources of linguistic knowledge and to perceive the inherent relationships between words and word groups. * This research has been supported by the 21st Century Center of Excellence Program “Framework for Systematization and Application of Large-scale Knowledge Resources”. The authors would like to acknowledge here the generosity of the Center. The first and second authors have been supported by Grants-in-Aid for Scientific Research from the Japanese Society for the Promotion of Science (research project number 19700238 to first author and 18500200 to the second). In addition, the authors wish to express their thanks to Professor Shun Ishizaki for permission to use his Associative Concepts Dictionary in this study. Copyright 2007 by Maki Miyake, Terry Joyce, Jaeyoung Jung, and Hiroyuki Akama 321 Although studies applying versions of the multidimensional space model, such as Latent Semantic Analysis (LSA) and multidimensional scaling, to the analysis of texts have been fairly fruitful, the methodologies of graph theory and network analysis are particularly suitable for elucidating the important characteristics of semantic networks. Recently, a number of studies have applied graph theory approaches in investigating linguistic knowledge resources (Church and Hanks, 1990; Dorow, Widdows, Ling, Eckmann, Danilo and Moses, 2005; Steyvers and Tanenbaum 2005; van Dongen, 2000; Watts and Strogatz, 1998). For instance, Dorow, et al (2005) utilize two graph clustering techniques as methods of detecting lexical ambiguity and of acquiring semantic classes instead …

Content truncated to 500 words