A clustering method and its application for keywords from literature search results

Currently, the articles search results from the literature database are only analyzed based on the external characteristics, which cannot reveal the related relationship between them. Articles returned to a user query are numerous and in a linear list, how to further classify these articles to allow users easily find related and needed articles is an important task in literature search. This paper proposes a new keywords clustering method for the literature search result clustering. In this method, keywords is extracted from the retrieved articles, the correlation between keywords is calculated by log-likelihood (LogL) ratio, and then the AP clustering algorithm is used for keywords clustering to form keyword clusters which is used for articles hierarchical classification. Finally, we take journal database of China National Knowledge Infrastructure (CNKI) as a data source of our experiment. The results show that the proposed method is feasible and available. It also indicates that the display of the hierarchical keywords classification results provided by this method is more effective comparing with the flat list of keyword classification from CNKI. ©, 2014, Wuhan University. All right reserved.