CredRank: Evaluating tweet credibility during high impact events

摘要

Twitter has evolved from being a conversation or opinion sharing medium into a platform to publicize information about high impact events. Now day's any events in the real world create a corresponding flood of tweets on Twitter, but not all content posted on Twitter is trustworthy for providing information about the event. The primary objective of this paper is to analyze the credibility of information in tweets corresponding to high impact events. The identification of content and sourced based features will be using regression analysis. Prominent content based features are number of unique characters; swear words, pronouns, and emoticons in a tweet, and user based features, for instance, number of followers and length of username. Furthermore, it reverts at implementing an algorithm to improve the trustworthiness and relevance assessment of micro blogs, using automated ranking techniques to assess credibility at the most atomic level of information on Twitter, i.e. at `tweet level'. It also includes URL and Hash Tags based ranking techniques which further improves the results and credibility of the algorithm. The preferred architecture utilizes the fundamental features of tweets to full extent and provides user with the most efficient information.