The Research Of Chinese Reviews Mining Technology And Applications In E- Commerce

As the important carrier of the opinion of the customers, Chinese product reviews are very significant for the potential customers to make the shopping decision and the manufacturer to improve the quality of products. As a result, this thesis hinges around improving the accuracy rate of sentiment polarity discrimination and does the analysis and research of the mining technology of Chinese product reviews.The thesis firstly elaborates the research status of attribute extraction and sentiment analysis respectively. From the development status, we can conclude that there are three problems in the mining technology of Chinese product reviews:the different importance levels of attributes is not put into full consideration, the degree adverbs can not be assigned with a weight if the abverbs are not in the list of basic degree adverbs and the extracted features are not comprehensive in the review mining techno logy based on the classifier.Focusing on above problems, this thesis designs mining method for Chinese product reviews combined with the weights of attributes and degree adverbs. The scheme discriminates the sentiment polarity from fine grit to coarsness. This method sets the weights of different attributes using the calculation of weights of attributes taking full consideration of degree of attention of the customer. This strategy can improves the accuracy rate if there are positive and negative polarity in the same sentence. This thesis also designs a method to calculate the weights of degree adverbs using semantics similarity based on lexicon of basic degree adverbs. This can improve the disadvantage that the way of calculating weights of degree adverbs, which uses matching method with lexicon of basic degree adverbs, can not set weights for all the degree adverbs and decreases the misdetection rate of the sentences of polarity discrimination. The thesis fatherly explores the efficiency of the designed method on multi-polarity determination. The simulation results show that the sentiment analysis combined the weights of attributes and degree adverbs can decrease the misdetection rate of the sentences of polarity discrimination and improve the accuracy rate of polarity discrimination. To improve the accuracy rate of the sentiment analysis combined with the weights of attributes and adverbs, this thesis designs the mining technology utilizing computational feature of significant review sentences. This strategy firstly generates the computational features of the significant review sentences using the sentiment analysis using the weights of attributes and degree adverbs and the computational features combined with general statistical features and basic statistical features are considered as the input of the classifier for the sentiment analysis of the whole sentences in coarse granularity. The simulation results indicate that the sentiment analysis combined with the computational features of significant review sentences improves the accuracy rate of the mining-technology combined with the weights of attributes and degree adverbs.At last, the thesis summarizes the research content of this thesis and outlooks the followed research direction.In a word, this thesis, which considers improving the accuracy rates of polarity discrimination as the ultimate goal, designs two mining methods of Chinese product reviews combined with the weights of attributes and degree adverbs and computational features of significant review sentences and uses the simulation to verify and analyze the efficiency of the methods on different datasets to improve the practical application ability of the methods.