A Part-of-Speech Based Sentiment Classification Method Considering Subject-Predicate Relation

Based on the topic and opinion classification, a tweet credibility analysis method is proposed to detect false information or rumors spreading on Twitter on and after the Great East Japan Earthquake. The credibility is assessed by calculating the ratio of the same opinions to all opinions about a topic identified by topic models generated using Latent Dirichlet Allocation. To identify an opinion (positive or negative) expressed in a tweet, a sentiment analysis is performed using a semantic orientation dictionary. However, the accuracy is a problem to identify the few false tweets. The accuracy of the originally proposed method was susceptible since the sentiment opinion of most tweets was identified negative by the baseline (namely Takamura´s) semantic orientation dictionary. Furthermore, specialty namely expertise of users was not considered. To cope with these problems, a method for extracting sentiment orientations of words and phrases was proposed considering user´s specialty / expertise degree or mark to each of the same / opposite opinion tweets. The effects of both improvements are proven by experiments using a large number of real tweets. Namely, in these experiments rum or tweets were detected more accurately.