Fine-grained Sentiment Analysis Based On Online Reviews

With the rapid development of the Internet and e-commerce in China,people are more and more keen on shopping on the Internet.The comments on the products contain a lot of valuable information.On the one hand,consumers can understand the reputation of the products through product reviews.In turn,the corresponding purchase decision is made;on the other hand,the manufacturer can find out the problem of the product through comments,thereby improving the product quality.Sentiment analysis,also known as opinion mining,emotion analysis,and sentiment orientation analysis,is a specific application of natural language processing.It is a process of extracting,organizing,and analyzing subjective texts containing emotions by means of data mining and machine learning.The results of sentiment analysis for online reviews provide potential consumers and businesses with the necessary decision information,so emotional analysis of online reviews is especially necessary.Traditional sentiment analysis generally divides text into two different types according to the emotional information expressed by the text.It can't help consumers to reduce the burden of information filtering at a more fine-grained level.More importantly,some comments are in the comments.The attributes of the evaluation object are not directly displayed in the statement,and their understanding may need to be judged by the context of the context.With the deepening of research,the current sentiment analysis can be divided into coarse-grained sentiment analysis and fine-grained sentiment analysis from the granularity of text analysis,and explicit sentiment analysis and implicit sentiment analysis from the perspective of attribute characteristics.This thesis has conducted in-depth research on these issues,and its main work and innovations are as follows:(1)Aiming at the quantitative requirements in fine-grained sentiment analysis,a method for establishing domain-level fine-grained sentiment lexicon is proposed.First,use the HowNet dictionary to match the common sentiment words in the field,remove the emotional words that are not related to the domain in the HowNet dictionary,join the network dictionary,and combine the attribute words,modifiers,and negative words to construct a fine-grained emotion dictionary in the form of five-tuples.At the same time,the different weighting effects of the negative words and the qualifiers on the emotional words are considered to facilitate the calculation of the emotional intensity score of each attribute when calculating the emotional intensity,so that the final quantitative score is more accurate.(2)Aiming at the characteristics of explicit features and implicit features in the commentary information,a multi-strategy method for mining explicit feature-viewpoint pairs is proposed,which improves the accuracy of extraction compared with the traditional feature-view method.At the same time,implicit attributes and emotional words are often not in pairs.Therefore,it is difficult to explore this difficulty.The implicit emotional words contained in the commentary are extracted,and then the combination of domain sentiment dictionary and machine learning is used to find Paired feature words.Finally,the explicit implicit emotions are combined and quantified to obtain the fine-grained emotional tendency of the product.It is proved by experiments that combining implicit emotions is helpful for the emotional analysis results.(3)In view of the traditional fine-grained sentiment analysis and multi-processing positive comments,the sentiment analysis method based on the pain point perspective is proposed,which pays more attention to the negative comments in product reviews,and gives the calculation method of pain point index to numerical information.Let consumers more intuitively understand the pros and cons of each feature of the product.Applying the method to the online reviews of different hotels,the visualized results give the commercial competitive advantages and existing problems of different hotels,and help the subsequent improvement of the business.Finally,the conclusions are given and the direction of further work is briefly discussed.