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.
