Research On Key Technologies Of Text Sentiment Analysis For Electric Business Reviews

Businessman can learn about the needs of customers and understand product inadequacies through the sentiment analysis of Electric Business Reviewsin timely,thanks to quickly adjust sales strategy.Therefore,the sentiment analysis of product reviews has enormous commercial value.However,it has many challenges in sentiment analysis due to the characteristics of short texts,weak normative,complex emotional semantics and so on.In order to solve these problems,This thesis focused on innovative research on aspects of phrase resource extraction,sentiment resource extraction,and sentiment classification technologies by making full use of the features of reviews.The specific work is as follows:1? Aiming at the issue of phrase resource extraction problem.This thesis researched how to efficiently extract comment phrases from product reviews.Based on the characteristics of E – commerce in the reviews,designing an IMEM(Iterate Multiwords expression Model)phrase extraction system by using an improved PMI(point-wise mutual information)iterative algorithm.The system can automatically extract high-frequency phrases in reviews rely on setting different threshold.2? Aiming at the shortcomings of construction methods for traditional sentiment dictionary,this thesis proposes an new sentiment dictionary construction method based on combination of statistics and rules.It consists of two parts of the extraction of basic sentiment words and sentiment phrases:(1)In order to solve the challenges of extracting the basic sentiment words from weak language norms,this thesis proposed a new method that by filtering based on part-of-speech(adjective,idioms)and splitting sentiment phrases to efficiently get it.(2)In order to solve the problem of sentiment word domain dependence,context dependence,modifier dependence,this thesis presented a method that combines statistical information and rules to extract sentiment phrases to solve these problems,and taking the camera corpora as the research object,verified it's value by a sentiment analysis method based on sentiment dictionary.3? Aiming at the difficulty of extracting sentiment features,this thesis proposed word feature optimization method based on combining phrases and sentiment weights to better represent text semantic information and sentiment information;Aiming at the problem that the high feature dimensions,data representation sparse in traditional vector space model,this thesis presented a new text representation method based on improved Word2 Vec and emotional weights.And the effectiveness of the method is proved by experiments.In addition,this paper also explores the influence of different word feature weights(word frequency,TF-IDF,boolean)on sentiment classification.