Research And Application Of Bus Opinion Analysis System Based On Sentiment Analysis

In the government work report,Premier Li Keqiang put forward the concept of"Internet+".After that,traditional industries are also actively exploring a road,that is,how they could achieve their own transformation through the Internet platform with the combination of online and offline,to create the value of the marginal effect.As a traditional industry,bus industry want to predict the passenger travel and help to make a decision that how to adjust line through collecting and analyzing the information of site,traffic,public opinion and so on.So,sentiment analysis based on the artificial intelligence plays an important role in it.At present,the research of emotion analysis has been relatively mature.But in the actual industry,the existing emotion analysis technology adopts the supervision-based way,which has high accuracy,but with a low portability,and high labor costs.Therefore,according to the characteristics of bus public opinion,this paper uses unsupervised emotion analysis technology and improves the traditional technology.The specific work includes the following aspects:1.This paper proposes an emotional dictionary expansion method based Word2Vec.The method combine the domain with semantic information of the words covering high use emotional vocabulary as much as possible.2.A new text representation model,RPFLO,is proposed.The model establishes the corresponding relationship between the emotion word and its evaluation object,and reveals the hidden semantic relation between the sentence orders in the long text.3.This paper proposes an event subject extraction method based on RPFLO model.This method improved the K means clustering algorithm,the common substring is used to realize the automatic selection of the initial class center.4.A similar topic clustering method based on improved Cure algorithm is proposed.First of all,the preprocessing of outliers improves the efficiency of the algorithm.Secondly,the inaccessible classes are introduced to terminate the clustering process automatically,which overcomes the disadvantages that traditional algorithms need to given clustering termination conditions(such as the number of clusters).Through the research,this paper realizes the optimization of the unsupervised sentiment analysis technology.On the basis of ensuring high transplantation and low labor cost,the analysis precision is improved greatly.