The Semantic Orientation Analysis Of The Web Review's Opinion

The network has already been recognized as the 4th media following newspaper, broadcast and TV, which becomes one of the main carriers to reflect people's emotion. Users issue their own opinions on a certain product in the network. It is worth considering for producers and potential buyers. The producers hope to know properties of their own product, and also know the competitors' properties of product. The potential buyers hope to have the suggestions from others' appraisal. But information of Web is too much. Only rely on artificial to browse Webpage through and analyze opinions is to be a time-consuming but less effective work. So, it is valuable to research the technology of opinion mining.Semantic polarity analysis and opinion mining is refer to utilizing the computer technology to analyze sentence or file automatically, and extract the theme or characteristic that users are interested in, and analyze its semantic polarity orientation (positive, negative or neutral) and intensity. Traditional information mining and information retrieval study are focus on the fact information that is expressed objectively, which is not suitable for the network review's opinion mining. The opinion is subjectivity, it needs the technology of mining opinion automatically, find out the semantic polarity orientation of the products' characteristic automatically in users reviews.At present, a lot of research institutions have already used the technology of The Natural Language Processing to do network review's opinion mining, which has made certain achievement. This text tries to utilize natural language processing method, continue to do further research to the semantic polarity orientation of Chinese sentence. This text has made some suggestions and some improvement of algorithms.According to the characteristic of the context polarity and some noun, this text solves the dynamic polarity problem through the method of giving polarity for these nouns. Considering the aspect of opinion mining, in order to solve the relationship of the theme word and polarity well and estimate the semantic orientation of each theme in sentences, this text draws the semantic model and dependency and makes a new algorithm which is named Polarity constriction and passing algorithm, establishes the theme opinion finally. This text carries on research to the exclamatory sentences and question sentences which makes certain achievement. Finally, this text compares the result of opinion mining automatically and the artificial result in the experiment; obtain the recalling rate and accuracy rate. Through the result of comparing with the data of other algorithms, it proves the validity of the Polarity constriction and passing algorithm. This text has solved some problem that other researchers have not resolved, and makes the Polarity constriction and passing algorithm. But this algorithm is unable to obtain the correct result to some other special sentence patterns. It needs further research.