The Use Of Sentiment Tag On Social Text Analysis

With Development of internet, social media becomes more and more important and domain. Blog, which is one of social media, is very famous to the society, while twitter which develop rapidly recent years also play a very important part. Meanwhile researcher are also very interested in this area. They study the classification of blog, the analyze the topic on twitter, they also predict things on all these platforms.Sentiment Analysis is also a very hot area recent year and it changes when time goes on. Now the main area has turned from the basic classification such as the classification of word, sentence and document, to more wide area such as topic analysis, prediction, and facet extraction and so on.In this paper, we focus on the use of sentiment tag which is the classification of certain word from some certain view. We use this feature in blog and twitter helping us with the blog classification, topic analysis and duration prediction.First of all, the sentiment tags are used to get the sentiment topic of blog which helps the blog classification. LDA model is used to get the sentiment topic meanwhile these topics are also needed to be filtered in order to select the topic which can classify the blog. Experiment result shows that this model performs well.In addition, in topic analysis area, sentiment tags are used to establish the edge of certain noun and make it to be a graph. And HITS algorithm is used to get the topic word for certain day considering the times of words as well. In this part, the sentiment of the tweets under certain topic word are also calculated in order to get the attitude of the public.At last, the duration of certain topic are predicted using the multi linear regression. The controversial Sentiment Model (CSM) is proposed to represent the controversial sentiment. Ultimately, a general model, Sentiment Impact Model (SIM) which can reflect positive, negative and controversial sentiment by adjusting certain parameter, is proposed to measure the sentiment impact to the topic. Experiment result shows that sentiment feature is beneficial to predicting the duration of the topic. SIM provides a better result than CSM and other baseline models.Above all this paper shows that sentiment tags can help us in classification, topic analysis and duration prediction area. In the social media background sentiment tags are are useful and can do more and more in future work.