Multi-Granularity With Multi-Classify Sentimental Analysis System

With the development of internet, many people put more time on internet for entertainment. The needs of internet are deepening, the internet has gradually become an indispensable role in life.In the beginning, people just can receive message from website; As the internet technology innovation, people was no longer a receiver but a sender, he become a node of internet. People can express their views, comments, and they also can share their interests, they are just used to show their feelings. In this context, sentimental analysis has become a hot topic in the academic.This paper discusses in detail some aspects of sentimental analysis, includes opinion classify, opinion mining and feature select etc. and do some work in some aspects:1. the design and implement of preprocessing modelReferred and researched the interrelated system and technology of preprocessing. Combined the feature of sentimental analysis, I designed the basic model of preprocessing. The model include feature selection and the computing of the weight of feature. Simultaneously, I propose some new features, for example, negative selection, syntax analysis, word etc. These features finally proved is effective for sentimental analysis.2. the design and implement of sentence-level sentimental analysisIn sentence-level I introduce the fusion of multi-classifier algorithm to solve the problem of sentimental analysis. This algorithm can adapt different corpus, so in average it can improve the accuracy of sentimental classify.3. The design and implement of Sentence-level sentimental analysis modelIn sentence-level, I introduce fusion of multi-classifier algorithm to solve sentimental analysis. This algorithm can adapt different corpus. Its average result can be effective improved.The result of experimental shows all of the three aspects can improve the sentimental analysis.