Features for Ranking Tweets Based on Credibility and Newsworthiness

We create a robust and general feature set for learning to rank tweets based on credibility and newsworthiness. In previous works, it has been demonstrated that when the training and testing data are from two distinct time periods, the ranker performs poorly. We improve upon this by creating a feature set that does not overfit a particular year or set of topics. This is critical for robust analysis of social media over time. In order to derive such features, we use the studies done on credibility perception of social media as well as the clues provided in past works in this domain. We also present new features that, to our knowledge, are more effective than the state of the art.