Application of TextRank Algorithm for Credibility Assessment

In this article we examine the use of TextRank algorithm for identifying web content credibility. TextRank has come to be a widely applied method for automated text summarization. In our research we apply it to see how well does it fare in recognizing credible statements from a given corpus. So far, research into use of NLP algorithms in credibility assessment was focused more on extracting the most informative statements, or dealing with recognizing the relation between claims within a document. In our paper, we use a collection of 100 websites reviewed by human subjects in regard to their credibility, therefore allowing us to check the algorithm´s performance in this task. The data collected showed that the TextRank algorithm can be used for recognizing credibility on the level of aggregated statement credibility.