Fake News: A Method to Measure Distance from Fact

Fake News is widely recognized as a security problem that involves multiple academic disciplines; therefore, solving the problem of fake news requires a cross-discipline approach where behavioral science, computational linguistics, mathematics, statistics and cyber security work in concert to rapidly measure and evaluate the factual content in any article. The model proposed in this paper relies on computational linguistics to identify and baseline characteristics of a fact-based narrative, and the distance measure between a news story and the original fact-based narrative.Once quantified the content can be used to tag news stories for further analysis. This additional tracking of the pattern spread of news can reveal differences from fact-based narratives since these narratives rely on a natural spread while their fake counterparts rely, in part on bots and trolls to saturate the news space. Finally, the metadata created in this measurement, tagging and evaluation process provides valuable inputs for mining purposes in support of provenance. Provenance in this case differs somewhat from the traditional data provenance of reputation analysis, this provenance examines the various sources, but in terms of the historical evaluations of the newly acquired metadata as applied to author and publication corpuses.