Study Of Microblog Sentiment Analysis Based On Semantic Feature

Sharing and real-time of microblog makes microblog platform not only have vast user base,but also produce massive microblog data,Microblog covers everything,including education,economy,science and technology,culture and so on.Processing and analysis of microblog data can mine hidden social value and economic benefit,timely understanding people's attention and emotional changes to a product,person and event,and provide a theoretical basis for decision makers in real time.Sentiment analysis of microblog is the most hottest research subject about sentiment analysis,and also one of the hottest issues about natural language processing.Research about Microblog Sentiment analysis is still in the primary stage and the range of application is limited in a small scope,so the further study is needed.Now,the most common method for microblog sentiment analysis is method based on sentiment dictionary and rule and based on machine learning.Sentiment word determines the microblog emotional bias,we must construct a complete sentiment dictionary to handle large-scale microblog data,but there is no effective method to construct a sentiment dictionary.Method based on machine learning need select features,but there is no effective feature can represent microblog.For the problem of no effective construction of sentiment dictionary,this paper present a construct method of sentiment dictionary based on Tongyici Cilin and microblog index system.In the paper,the four basic sentiment dictionary is extended to a more perfect dictionary by Tongyici Cilin,and microblog index system in million level is constructed and sentiment tendencies is judged through the dictionary together with PMI formula.For the problem of feature selection,this paper selects effective features through different feature combination including sentiment dictionary.Experiment results show that,the sentiment dictionary constructed in this paper is more complete,compared with the four basic sentiment dictionary,performance of sentiment analysis is improved effectively.Feature combination in this paper can solve feature selection problems in machine learning.This study improves the quality of sentiment dictionary and makes up for the lack of feature selection,and provides new method and idea for sentiment analysis of microblog.