Abstract— There is the research in fashion that expresses the text analysis result with network structure between words in recent years. Clarifying the relation of the words is important for the future legacy text mining technologies such as a derivation of the conceptual meaning between words and its relationship, new evaluation indexes for degree of similarity between documents, and a visualization of the relationship. In such text network analyses domain, a method of node evaluation is not defined clearly, so far. For this background, the intensive comparative evaluation has been made with three typical indexes in network analysis which are degree centrality, closeness centrality, and betweenness centrality. We have made a conclusion that the betweenness centrality marked the best result.

Index Terms— Centrality, graph, meaning, network analysis, network text analysis, ontology, text analysis.

Koji Tanaka is with the Graduate School of Business Sciences, The University of TSUKUBA, Tokyo, 3-29-1 Otsuka, Bunkyo-ku, Tokyo 112-0012 Japan and with Hitachi Government & Public Corporation System Engineering, Ltd, 2-4-18 Toyo, Koto-ku, Tokyo, Japan 135-8633 (e-mail: tnkouji@gssm.otsuka.tsukuba.ac.jp). Kazuhiko Tsuda is with the Graduate School of Business Sciences, The University of Tsukuba, Tokyo, 3-29-1 Otsuka, Bunkyo-ku, Tokyo, Japan 112-0012 (e-mail: tsuda@gssm.otsuka.tsukuba.ac.jp).
Masakazu Takahashi is from Graduate School of Innovation and Technology Management, Yamaguchi University, 2-16-1, Tokiwadai, Ube, Yamaguchi, Japan 755-8611 (e-mail: masakazu@yamaguchi-u.ac.jp).

Cite: Koji Tanaka, Masakazu Takahashi, and Kazuhiko Tsuda, " Comparison of Centrality Indexes in Network Japanese Text Analysis," International Journal of e-Education, e-Business, e-Management and e-Learning vol. 3, no. 1, pp. 37-42, 2013.