Topic clustering from selected area papers

초록

An extracting method of research trend in the field of a computer network which contained in the published papers of related conferences is presented. A topic is defined as a subset of vocabulary and the interest of the topic is represented as 'saliency'. The saliency, the degree of topic interest is measured as a product of joint distributions of vocabularies which consist the topic. To reduce the computational burden, clustering and selection procedures of vocabularies are applied before actual topic grouping. Two experiments: 1. Research trend analysis and 2. Topic correlation analysis of conferences has been performed. The leading 24 conferences related to the computer networks which are held during 2009-2010 are exploited. The experimental results show the validity of the presented method.

키워드