Opinion Community Discovery Based On Sentiment And Region Of Web Comment

With the development of Web2.0, the Internet has become an important channel for the netizens to express their sentiments and opinions. Currently, more and more netizens public their opinions about social occurrences and phenomenon by means of such tools as blog, forum, the news sites of comment and groups. More and more contents generated by netizens on the web can reflect public opinions to a great extent. For the government agencies, public opinion is very useful in guiding their public information and propaganda programs and occasionally for helping in the formulation of other kinds of policies. So, analyzing the contents of the public opinion is very important, and web comment is an important data source for opinion mining.There have been a lot research on web comment analysis, but current researches in this domain mainly focus on sentiment classification which classifies netizens’sentiments into only positive, negative, and neutral. For obtaining richer sentiment, an approach for discovering opinion community based on sentiment and region of netizens is proposed in this thesis. A clustering technique is applied to partition netizens’sentiment into different opinion communities based on sentiment similarity, the concept of longest sequential sentiment phrase (LSSP) is defined and a LSSP mining algorithm is proposed to extract the typical opinions for every community. Moreover, the region distribution for every opinion community is shown in Chinese map.The experiments show that approach proposed in this thesis can provide more information for web opinion analysis comparing with related work.