Research On Fine-grained Sentiment Analysis Base On Chinese Micro-blog

At the start of this century, text sentiment analysis gradually becomes the researchfocus in the information extraction field, and gains more and more attention as well.Especially with the popularization of Web2.0technology, microblog obtains greatlove from the netizen due to its features like brief message, convenient publishing andreal-time interaction. The netizen, previously as simple information receiver, thushave turned into the main makers of Internet content. Meanwhile, with the continuousimprovement and maturity of natural language processing and machine learningtechnology, sentiment analysis of subjective short text like microblog becomespossible, and is gradually extensively used.In China, currently, sentiment analysis study of Chinese microblog mainlyfocuses on emotion polarity judgment, such as the analysis of whether the microblogexpression is positive or negative. This type of research has gained a certainachievement and been extensively used, however, with the application in a deep-goingway, the user hope that they can acquire more delicate emotion that the microblogexpresses to understand the user’s attitudes towards something. At this moment,traditional research methods of microblog sentiment analysis are hard to completelysatisfy the needs. Although researchers are trying to do some studies aboutfine-grained emotion analysis, the results are not satisfied. Therefore, this paperstudies further improvement on the accuracy and practicability of sentiment analysisthrough fine-grained emotion analysis method of Chinese microblog and explorationof new research thought and methods.This paper expands its study mainly by directing at key technology offine-grained emotion recognition of Chinese microblog. It analyzed research difficulties and emotion expression features of Chinese microblog, and put forward akind of multi-strategy integration analytical method based on emotional vocabulary.Firstly, in the premise of massive tests collection, we put forward iterative na veBayesian classification algorithm in the emotional and unemotional classification ofmicroblog. The classification of this kind of algorithm is emotional microblog text,we further analyze delicate emotion it expresses, including anger, disgust, fear,happiness, like, sadness and surprise. Novel textual features vector representation andweight calculation to quantize emotional microblog text are used, and7kinds offine-grained emotion classification have been conduct based on SVM and KNNrespectively, which makes fine-grained emotion analysis of Chinese microblog.Finally, with the help of noumenon Dalian University of Technology, Sinamicroblog as the experimental subject, and through experimental comparative studyof multi-strategy integration emotion analyzing alternatives and traditional analyzingalternatives, the result indicates analytical method based on multi-strategy is superiorto the single traditional sorting algorithms. While in multi-strategy analytical method,NB&KNN is superior to NB&SVM.The main contribution of this paper is:1) We propose Navie Bayesian Classification Algorithm Based on Iteration,which can improve the classification performance in the case of lacking futureknowledge.2) Here we enrich and extend the emotional ontology library of Dalian Universityof Technology. It will be a great help to enhance the accuracy rates of the fine-grainedemotion recognition.3) A new method of features vector presentation and weight calculation isproposed. It plays a crucial role as a way of dealing with the curse of dimensionality,reducing computational complexity and improving the algorithm performance.4) We put forward an analytical approach of fine-grained emotion recognitionbased on multi-strategy integration. The result indicates analytical method based onmulti-strategy is superior to the single traditional sorting algorithms. In the meantime,it provides more ideas and strong theoretical basis for the fine-grained emotion analysis.Through fine-grained emotion recognition analytical investigation of Chinesemicroblog, significant decision basis will be provided for promoting e-commercedevelopment, organizing institutional opinion survey, and monitoring public opinion.However, the accuracy of fine-grained sentiment analysis still has large room forimprovement.