Fine-grained Emotion Analysis Based On Online Reviews

With the rapid development of Internet and electronic commerce,the network provides convenient ways and platforms for information dissemination,communication and communication for enterprises and consumers.Consumers are usually evaluated on the Internet based on their satisfaction after online shopping,and these comments express the emotional tendencies of consumers.The emotional tendencies contained in these reviews are very helpful to consumers,businesses,and relevant government departments.For consumers,they can provide the fact of products at the time of buying;for businesses,they can help them realize the lack of quality and management;for relevant government departments,they can be used as references for their supervision.A large number of online reviews contain a lot of emotional tendencies,and research and mining of useful information is a problem we need to solve at present.This article is the study of emotional tendencies,especially fine grained emotional tendencies and intensities,based on real online reviews.This article is the study of emotional tendencies,especially fine grained emotional tendencies and intensities based on real online reviews.This paper first introduces the theoretical knowledge of text mining and emotional analysis,including the model of Chinese word segmentation,text feature extraction and text classification algorithm.On these bases,the paper analyses the complexity of online reviews,and the screening of useful comments is proposed to reduce the interference of unused comments on subsequent affective analysis.It is proved by the experiment that the feature selection method IG method combines with SVM classifier is the most effective choice is to select the useful reviews.Secondly,this paper summarizes the relationship between the feature words and form of the evaluable words,as well as rules of composition evaluation,and then put forward a mixed feature extraction algorithm for view based on dependency parsing,which can more accurately mining the features and opinions of online reviews in a fine-grained level,ensuring the accuracy and recall evaluation of collocation extraction.Then,according to the features extracted from the view of online comments,this paper uses the way to judge the emotional tendency and emotional strength based on Hownet method.At last,this paper realizes the sentiment oriental analysis of reviews information on fine-grained level,compared to coarse-grained,this method is more accurate and effective and comprehensive to qualify the emotion of each attribute of commodity,can provide better support for consumers,businesses and relevant government departments to make decision.