Research On The Chinese Microblog Sentiment Based On Sentiment Lexicon

With the rapid development of computer network technology,Internet has import role in our daily information media,so that online documentation has become an indispensable part of people's daily life.Microblog,as one of the most influential social media,is well received by the public.However,emotional information carried by views has the practical significance in various application,which is realized in micro-blog marketing,brand promotion,customer relationship management,public opinion monitoring and so on.Therefore,the emotional analysis as an important part of Natural Language Processing in recent years.Especially,favored by many scholars,micro-blog's emotional analysis has become the focus of the current study.Emotional analysis is mainly to determine the emotional tendency of micro-blog text,that is positive,negative,neutral.The construction of emotion dictionary and the recognition of emotion tendency are the important parts of text sentiment analysis.In this paper,we take the high quality affective dictionary algorithm as the research object firstly,and further study the construction of the network dictionary.On the one hand,makes a selection and arrangement of open source emotional dictionaries which is relatively authoritative to get basic dictionary;on the other hand,it is focused on the expansion of the network emotion dictionary.According to the difference of the micro-blog training,this paper puts forward the CHI-Order algorithm based on the labeled corpus and the CO-PMI method that does not need to label micro-blog experiment corpus manually.Besides,in experiment part,on the basis of micro-blog data,this paper mainly compute emotion value based on mentioned above.In order to complete the reasonable selection of effective emotion words,we set the threshold especially,thus constructing a Chinese emotion dictionary for micro-blog.At last,the paper designs the experiment to verify the validity of the emotion dictionary.The experimental data is related to the COAE evaluation copus,and results verified the feasibility of the method in our paper.The experiental results show that the accuracy of CHI-Order and CO-PMI algorithms are 81.89% and 74.14%,respectively,and the results are well received.In addition,In order to improve the accuracy of micro-blog emotion recognition,the method based on relationship between sentences is proposed.Apart from the basic vocabulary,pos features,emoticons and other features,the method took account for effect of the characteristics of conjunctions,the sentence features to emotional expression.The experimental results show that the method has achieved a rate of 80.49%,and the experimental results are higher than those of the general method.