Research On Sentiment Analysis Method And Optimization Of Electronic Commerce Comments

With the e-commerce market is developing rapidly,more and more customers shopping online,and online sales have also become the focus of the manufacturers sales.A sentiment analysis of the product reviews can provide an important reference for consumers to buy products and businesses to adjust sales programs.This paper takes the commentary of the Media's water heater as an example to study the sentiment analysis method of the short text which have the typical of text short,partial colloquial,rich emotional semantics and so on.The sentiment analysis method is optimized from the aspects of stop list,emotional dictionary and document frequency,and achieve good results.Details are as follows:In the process of text preprocessing,aiming at the problem that the existing vocabulary is not suitable for the sentiment analysis of the product reviews,paper constructs a stop list of this field.This article first statistics the frequency of the words which has been segmented,and then manually extract the words which has a high frequency but makes no sense of electricity supplier reviews sentiment analysis as a stop word.Finally,the extracted stop words are synthesized into a stop list.The validity of the stop list in the field of electricity supplier product reviews is proved by comparing with the traditional stop lists.In order to solve the problem of the non-applicability of the public emotional dictionary in the sentiment analysis of the electricity supplier product reviews,the paper constructs the emotional dictionary associated with this field.Paper first manually extracts part emotion words as benchmark words from the water heater comments,then calculate the semantic similarity between other words and benchmark words.The word will be added in the emotional dictionary which consisted with the benchmark words if it's value of the semantic similarity was higher than the threshold.The Effectiveness of the emotional dictionary constructed in this paper is verified by comparing the classification performance of the emotional analysis with several open emotional dictionaries.For the feature dimension is still relatively high after text preprocessing,paper proposes to combined with document frequency and TF-IDF to feature selection and feature weighted to solve the problem of high feature dimension and data sparse.At the same time,the paper also proposes to increase the feature of text length to solve the unbalanced distribution issue of the length between the positive and negative comments.At last,paper uses stop list to carry out the stop words and uses emotional dictionary to Chinese word segmentation,then integrates document frequency and TF-IDF and text length characteristics to construct ESDTL model of sentiment analysis for electronic commerce comments.The paper takes the comments of water heater for Jingdong as the corpus,takes support vector machine as the core,uses the ESDTL model to carry on the sentiment analysis experiment,the experimental result has proved that the proposed method can reduce the text dimension while greatly improving the accuracy of the sentiment classification,and obtains the good effect in the sentiment classification of the electronic commerce.