A Hybrid Approach for Credibility Detection in Twitter

Nowadays, microblogging services are seen as a source of information. It brings us a question. Can we trust information in a microblogging service? In this paper, we focus on one of the popular microblogging services, Twitter, and try to answer which information in Twitter is credible. Newsworthiness, importance and correctness are the dimensions to be measured in this study. We propose a hybrid credibility analysis which combines feature based and graph based approaches. Our model is based on three types of structures, which are tweet, user and topic. Initially, we use feature based learning to construct a prediction model. In the second step, we use the results of this model as input to authority transfer and further refine the credibility scores for each type of node. The same process is used for measuring each of the dimensions of newsworthiness, importance and correctness. Experimental results show that the proposed hybrid method improves the prediction accuracy for each of these credibility dimensions