Rumor Detection on Twitter using Extracted Patterns from Conversational Tree

These days, Twitter social network is one of the main platforms for getting news, among people over the world. This is because of the high volume of data generated by this social media, which makes Twitter up to date with news and information. Nevertheless, the existence of invalid information over the social network makes the users unhappy and also arises some problems in the real world, particularly in the crisis. To overcome these problems and other possible issues, automatic detection of rumor on Twitter must be taken into account. Despite such issues, in this paper, rumor detection in Twitter is studied. In this paper, the rumor is validated by considering the user's feedback, as the source data for rumor study. In our proposed method, pattern recognition and its analysis of the user conversational tree in Twitter is studied. These recognized patterns feed into as features for training a classifier for rumor detection. The model for training a classifier is an Extreme Learning Machine and its extension. The dataset for experiments of our method is the standard dataset of SemEval-2017 Task 8. Experiments of our proposed method with respect to competitor methods in rumor detection show that our method outperforms the state of the art methods.