@Twitter credibility ranking of tweets on events #breakingnews

Abstract

Large amount of content is generated on online social net- working and micro-blogging services daily; Twitter is one such micro-blogging service. Twitter has evolved from be- ing used for conversing with friends and expressing opinions into a medium to share and disseminate information about current events. Events in the real world creates a corre- sponding spur of tweets in Twitter. In this paper, we an- alyzed tweets corresponding to fourteen major news events of 2011 around the globe. We empirically show that the properties of information di usion (via retweets, and URLs) on Twitter di ers during crisis and non-crisis events. Us- ing supervised machine learning and relevance feedback ap- proach, we show that ranking of tweets based on Twitter features can aid in assessing credibility of information in messages posted about an event. We found that both mes- sage and source based features help in predicting the rank of the tweets. The performance of ranking algorithm was signi cantly enhanced by using reranking strategy as it pro- vided context speci c (unigrams) features to the algorithm. To this best of our knowledge, this is the rst work to study credibility of content on Twitter at the tweet level and ex- ploring an automated ranking framework to predict rank of tweets according to their credibility.