Web-based statistical fact checking of textual documents

Abstract

User generated content has been growing tremendously in recent years. This content reflects the interests and the diversity of online users. In turn, the diversity among internet users is also reflected in the quality of the content being published online. This increases the need to develop means to gauge the support available for content posted online. In this work, we aim to make use of the web-content to calculate a statistical support score for textual documents. In the proposed algorithm, phrases representing key facts are extracted to construct basic elements of the document. Search is used thereon to validate the support available for these elements online, leading to assigning an overall score for each document. Experimental results have shown a difference between the score distribution of factual news data and false facts data. This indicates that the approach seems to be a promising seed for distinguishing different articles based on the content.