Research On Sentiment Analysis Of E-commerce Online Reviews Based On Text Mining

The emotional attitude contained in online reviews of e-commerce is of great value to enterprises.A large amount of review information contains consumers' evaluations of all aspects of products or stores.This information is not only an important reference for other potential consumers to decide whether to purchase products,but also product manufacturers to grasp consumer preferences and consumption trends,and quickly find user needs And improve the direction,enhance the core competitiveness of products,and continue to develop an important source of information for new products that meet the expectations of consumer groups.At the same time,the current technological development has long been different from the past,and products are iteratively updated.Old products are constantly being eliminated and new products appear.In this increasingly competitive market environment,stable growth can only be achieved by ensuring that products meet customer needs.However,seeking product improvement through traditional methods such as user feedback surveys can no longer fully meet the needs of the moment.By fully mining and utilizing the huge amount of information in the era of big data,we can better grasp the key.Based on the theory of consumer sovereignty and purchasing decision theory,this thesis takes online review texts of mobile phone products on e-commerce platforms as the research object,and uses methods such as web crawlers,Word2 vec word vectorization,and emotional dictionaries to construct the evaluation of online reviews of mobile products.Indicators,and verify the effectiveness of the method through principal component analysis and multiple linear regression analysis.By combining evaluation indicators,mining mobile review texts,analyze consumer demand for products and make relevant recommendations for companies.The main research contents of the thesis are as follows:First of all,the paper explains the definition of text mining,sorts out the three development stages of text mining,and summarizes the general steps of text mining;after expounding the relevant theoretical basis,sorting out and analyzing the past scholars 'research on e-commerce online reviews,text sentiment Based on the analysis and research and the research results of e-commerce online review research based on sentiment analysis,it is clear that the paper takes e-commerce online review text as the research object,and through digging the diversified emotions within the text and quantifying the indicators,it can guide the production of the company's products.Purpose: Focusing on the relevant technical methods for the realization of the thesis research,the thesis introduces the relevant principles and implementation steps of text word segmentation,word vector tools,Kmeans clustering,principal component analysis and multiple linear regression analysis.Secondly,in order to realize the quantification of the emotional information in the e-commerce online review text,the paper crawled the online review text of 52 mobile phones in the mobile phone hot list of JD.com,and organized the traditional dictionaries such as How Net Dictionary and NTUSD Dictionary into the original text of the paper.Dictionary,using the Word2 vec model to train 52 mobile phone review texts processed by Jieba word segmentation to obtain word vectors.Combining Word2vec's similar word function and Kmeans clustering,the original dictionary is expanded to obtain the final dictionary of the paper,including the emotional word dictionary,Attribute word dictionary,degree adverb dictionary,negative word dictionary;through dictionary assignment,short sentence usefulness screening and stipulating five types of sentiment value calculation rules,Python programming is used to realize the quantification of sentiment and attitude;based on sentiment score,introduction to the paper The process and meaning of the construction of four evaluation indicators,including: satisfaction index,attention index,degree of improvement index and emotional variance index,combined with the negative evaluation rate of 52 mobile phones,using principal component analysis and multiple linear regression analysis The regression equation model was established to verify the validity of the evaluation index.Finally,through the four evaluation indicators constructed in the article,they are applied to the 52 mobile phone products in the Jingdong Hot List,and the evaluation indicator data is visualized.The visual display and analysis are carried out from the perspective of 52 mobile phones as a whole and from the perspective of high,medium and low-end mobile phones.,And combined with the index data to analyze the key attributes of mobile phone products such as experience attributes,battery attributes,packaging attributes,body appearance attributes,and camera attributes,and further pointed out the root causes of negative reviews.Related suggestions are given in terms of marketing strategy,in order to produce certain practical significance and value to the enterprise.