Identification and ranking of event-specific entity-centric informative content from twitter

Abstract

Twitter has become the leading platform for mining information related to real-life events. A large amount of the shared content in Twitter are non-informative spams and informal personal updates. Thus, it is necessary to identify and rank informative event-specific content from Twitter. Moreover, tweets containing information about named entities (like person, place, organization, etc.) occurring in the context of an event, generates interest and aids in gaining useful insights. In this paper, we develop a novel generic model based on the principle of mutual reinforcement, for representing and identifying event-specific, as well as entity-centric informative content from Twitter. An algorithm is proposed that ranks tweets in terms of event-specific, entity-centric information content by leveraging the semantics of relationships between different units of the model.