Analysis And System Implementation On Users' Comments Of Electronic Commerce Systems Based On Text Mining

Text mining and emotion analysis techniques play an important role in e-commerce.Anal-ysis and emotional mining of users' comments of electronic commerce systems not only provide enterprises with accurate data statistics,advertisement push and merchandise ranking,but also provide valuable reference for users to purchase the corresponding commodities.At present,the Chinese emotional dictionary is imperfect,and results of the dictionary-based emotion analysis method is lack of accuracy because it based only on the frequency of emotional words.This thesis analyse and implement a system on users'comments of e-commerce systems based on text mining.The research has important theoretical and practical significance.Targeted at the commentary texts of the e-commerce website,this thesis studies the method of emotional analysis based on the dictionary from the aspects of the extraction of emotional words,the construction of emotional thesaurus and the calculation of emotional values.The main work of the thesis includes the following aspects:(1)Aiming at the shortcomings of over generalization and weak adaptability of existing emo-tional dictionaries,this thesis proposes a method to extract emotional words based on depen-dency syntax.By analyzing dependency syntaxes of historical commentary data of Taobao,this thesis excavates dependency relations which includes the user's emotion tendency and builds a rule table.According to the rule table,this thesis extracts emotional words from a large number of historical evaluation corpus.After emotional words extraction,this the-sis annotates each word with an emotional value using semi-automatic method to build an emotional dictionary which is suitable for the field of electric business.(2)In order to solve the low accuracy problem of current dictionary-based emotion analysis method which calculates emotional value only based on the number of positive and negative emotion words,this thesis analyzes the influences of negative words,degree words and conjunctions in detail and establishes corresponding dictionaries.This thesis then proposes a method to compute the emotional polarity of the text comprehensively to improve accuracy.(3)Based on the above-mentioned research,archetype of user comments analysis system is designed and implemented in this thesis.By testing the system through a large number of commentary data of Taobao's mobile phone products,it obtains more than 90%accuracy,which verifies the feasibility and accuracy of the method.