Research On The Emotional Analysis System Of E-commerce Commodity Evaluation

With the rapid development of e-commerce,online shopping has becomes an indispensable part of our daily life,Accompanied by a large number of product evaluation data,which not only help consumers to buy appropriate products,but also as a business reference to improve various services.Therefore,the emotional analysis on evaluation data of online products becomes more and more important.The research topic of this paper is “Research on the emotional analysis system of e-commerce commodity evaluation” The main purpose is to make use of the increasingly developed computer technology to conduct emotional tendency research on the large-scale logistics-related evaluation of online products and investigate the consumers' opinions and attitudes toward the various logistics suppliers in the store.This paper first makes a brief introduction to the background of the topic selecion,the significance of sentiment analysis and the research status at home and abroad,and then explains the general steps of sentiment analysis.This article involves the following main points:1.In order to get more pure data,it is necessary to carry out data deduplication and data cleaning from initial evaluation data.2.Chinese word segmentation for evaluation data on logistics.This paper first introduces the traditional Chinese word segmentation method briefly,then according to the fact that most of the online product evaluation data have fewer characters belonging to the category of short text,we select the feature tagging method and feature template suitable for short text,reusing the dictionary of specialized terms in logistics and combining with the reverse maximum matching algorithm based on dictionary tree to get the Chinese word segmentation method which is applicable to the evaluation data of logistics in online product evaluation.3.This paper improves the traditional algorithm information gain method which is often used for feature selection in text classification.First,we remove the "useless words" to reduce the influence of low-frequency features,and then use the document frequency of feature items in each class to improve the imbalance between classes.Through these two means to improve the traditional information gain algorithm.4.We use SVM and Naive Bayes classifier respectively to test the improved feature selection algorithm and the traditional feature selection algorithm.The experimental results obtained are as follows: under the same experimental conditions,the naive Bayesian classifier is used to classify the data using the traditional information gain feature selection algorithm,the calculated F-value is 0.8573;For the improved information gain algorithm,its F-value is 0.8786;Similarly,using the SVM to classify the two algorithms,it is obtained that the F-value calculated by the traditional information gain algorithm is 0.9002;The F-value of the improved information gain algorithm is 0.9292.The experimental results show that the classification effect of SVM is better than that of naive Bayes;The improved information gain algorithm performance is better than the traditional information gain algorithm.