Research on Topic Detection Strategy Based on Extension of Comments and HowNet Lexeme
Yun Liu, LI Xiao-xian, Bing Zhao · 2014
As a product of Web2.0, micro-blog is developing rapidly these years. More and more information spread on the micro-blog because of its high speed and convenience, social hotspots and news events included. As a result, discovering, extraction and analyzing information become researching hotspots. By studying micro-blog text and long text cluster, this article draws a conclusion that traditional cluster algorithms cannot be used to discover topics because of the length of text. Therefore, this article proposes a solution which is based on the extension of the comments and HowNet lexeme. By this method, the short text and diversified expression can be overcome. Finally, the simulation results show that the proposed algorithm would significantly diminish the bad effects which are the results of short-text and improve the accuracy of clustering results.