Measuring Semantic Relatedness by Wikipedia Revision Information

• Main Authors: Wen-TengYang, 楊文籐
• Other Authors: Hung-Yu Kao
• Format: Others
• Language: en_US
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/25508215374684132603

碩士 === 國立成功大學 === 資訊工程學系碩博士班 === 98 === Nowadays, Wikipedia is an accurate, fashion, huge and wiki-based encyclopedia on the WWW. Simultaneously, Wikipedia is invaluable resource for research work because Wikipedia have many useful properties to enrich. Those properties contain high dense links, live update, URL identification for concepts and complete revision history, etc. In this paper, we deal with the articles which the words represent in Wikipedia. Moreover, each Wikipedia article represents an individual concept and simultaneously contains other concepts which are hyperlinks of other articles in its content. Namely, the semantic relatedness between two articles is also the semantic relatedness between two words. Therefore, we propose an Editor-Contribution-based Rank (ECR) algorithm for ranking the concepts in the content of all revisions of article and take the ranked concepts as a vector to represent the article. ECR ranks those concepts depending on the relationship between concepts and the editors. We classify four types of relationship which behavior of addition and deletion maps to appropriate and inappropriate concepts. The experiment shows our method is better with the development of Wikipedia and gets 4.4% improvement over previous methods which calculate the relatedness between two articles.