Ranking Assessment of Event Tweets for Credibility

Online social network services have become a popular web activity to establish online social relationship among the people all around the world. These online social networking services allow its users to share opinions or posts on any high impact events in the world. The primary task of these services is to sort out credible posts and provide credible information about the event to the users. In this paper, the focus has been on Twitter, a rapidly growing micro blogging platform, which provides a large amount, diversity and varying quality of content. As Twitter is open to all, it emerged as an excellent means to disseminate information to a large user community in the shortest time. Due to very open, uncontrolled nature, Twitter has become vulnerable to incredible information from malicious and credulous users. Consequently, it is important to formulate sophisticated methods for analysis of credibility and relevance for ranking tweets. In this paper, tweets of an event posted by users have been collected and allowed to perform annotation process on those tweets by three human annotators to assess the tweet credibility. In order to provide ranks to the tweets according to the features, content based features are extracted. The performance of the ranking strategy used to rank tweets according content based features has been enhanced by the re-ranking strategy which uses context specific features such as event specific words to re-rank the tweets. After re-ranking tweets, an evaluation has been carried out by NDCG metric to measure the accuracy of re-ranking strategy.