Web Public Opinion Orientation Analysis Based On Technique Of Sentiment Lexicon Extension

With the advent of the age of web2.0, the way of people using the network has significantly changed. People had been accustomed to passively consulting the information they needed previously. But now, more and more people take the initiative to express their views and opinions in the blog, forums, the news sites of comment and groups. The number of participation is greatly increased. The network has becoming one of the important instruments of reflecting public opinion. Due to the discrepancy in people’s positions, the point of departure, personal situation and preferences, there must be a lot of difference in people’s belief, attitudes, opinions and the propensity of sentiments. This difference is especially seen in forums, blogs and other internet media of reflecting grass roots’views. Therefore, finding the network public opinion, users’point and mining the propensity of viewpoint have become a new hotspot.Opinion mining is a kind of techniques which can mine the advice from the forum, discussion groups, news groups, blog, etc. Based on the theoretical research of opinion mining, domestic and foreign scholars had applied it to the relevant areas, and successfully developed a number of prototype systems and commercial products. From the point of view of application, it is fruitful on the application of e-commerce. However, it is rarely effective on the hot topic for online views and opinion mining on comments.In this thesis, for the characteristic of simple and a large number of public comments, a special crawler is firstly designed to climb from the viewpoint of netizens, and then e the emotional lexicon is extended using the resources HowNet and NTUSD Lexicon, and a new lexicon with the degree of emotional tendencies is built. Based on the expansion of the emotional lexicon, a semi-automatic analysis system is developed for public opinion. The system offers help to classify the comments and tendentious provided by users, which reduces the workload of manual sorting and accelerates feedback cycle of calculating the attitude of users about related topic.For semi-automated analysis system of network public opinion, the experimental results show that:the system achieved a good outcome of the trial.