Research On Sentiment Analysis Method For Chinese Reviews

With the development and popularity of online social media such as forums,micro-blogs,micro-letters and so on,a large number of opinions and speeches containing personal subjective feelings have appeared on the Internet.These comments have great commercial value and social value.By analyzing the subjective feelings contained in these comments,we can understand the public opinion about an event or views about product,not only can help potential users make decisions,but also help control public opinion.However,these massive comments can not be processed manually,but text sentiment analysis technology provides the possibility for rapid and accurate mining of emotional information.Despite the rapid development of sentiment analysis technology,there are still many methods to be improved and optimized.Therefore,after fully studying the theory of sentiment analysis,this paper divides the study of text sentiment analysis into coarse-grained and fine-grained parts,and improves and optimizes the traditional technology based on the Internet e-commerce commentary corpus,and makes the effect of emotion mining better.In coarse-grained text sentiment analysis,this paper classifies the commentary sentences as a whole,that is,according to the emotions contained in the commentary sentences.Feature selection plays an important role in sentiment classification.This paper study the CHI algorithm which has better effect on feature selection.Aiming at the defect of word frequency,this paper improve it by combining the particularity of sentiment analysis.This paper propose the word frequency computation quantity CHI_freq and the sentiment parameter ?,and combine them with the traditional CHI algorithm to form the IM-CHI feature selection algorithm.Experimental results show that the improved algorithm is better than the traditional algorithm.In fine grained text sentiment analysis,this paper extract the commentary information in the comment sentence.Commentary information refers to several pairs of "subject words-sentiment words" in a sentence,the subject words are the concrete evaluation objects and the modification objects of affective words.This paper extracts comment information based on CRF model,and constructs an emotional dictionary to optimize the extraction results,so as to make the extracted "subject words-sentiment words" more comprehensive and accurate.The experimental results prove the feasibility of the method.