Knowledge discovery from Chinese internet public opinion short text

Shengluan Hou, Lei Liu, Cungen

Abstract

Internet is becoming a spreading platform for the public opinion. Internet public opinion(IPO) discovers hot topics and spreads rapidly. It is vital to analyze the IPO and grasp their trends correctly and timely. A great deal of IPO text is short text which is oral and exits out-of-vocabulary words, such as Weibo text, products comments and so on. In this paper, we propose a novel semantic-based method of knowledge discovery from Chinese IPO short text. This method has two parts: Knowledge Discovery from Chinese IPO short text Language(KDIPOL) and parser of KDIPOL. An extensible context-free grammar(ECFG) is presented to describe the IPO short text. We adopt semantic constraints in ECFG to solve context-sensitive and ambiguous problems of Chinese. Then we design and implement a parser of KDIPOL, which provides a running platform for KDIPOL parsing and outputs frame-based knowledge representation form. Experimental results show the high performance of our method.