Identifying sentiment patterns of BBS reviews based on associateve memory model

摘要

BBS is popular online forum, which contains a great wealth of knowledge about private opinions and sentiments. Because the information in BBS is mess, it is difficult to identify this useful knowledge. Taking advantage of the functions of bidirectional associative memory, the paper presents a novel method to identify sentiment patterns of BBS reviews. We call it ISPBAM (Identify Sentiment Patterns based on BAM). It can acquire the syntax pattern of unnormal sentences in BBS reviews and identify the sentiment orientation of them. So it combines two functions of sentiment classification and polar terms recognization. But differ from simplex sentiment classification or polar terms recognition, this method can identify sentiment patterns without constructing linguistic resources. The experiments are done for BBS reviews about recent Chinese Spring Festival Gala Evenings. The results show the proposed method is feasible, and more powerful than former methods.