FactRank: Random Walks on a Web of Facts

Fact collections are mostly built using semi-supervised relation extraction techniques and wisdom of the crowds methods, rendering them inherently noisy. In this paper, we propose to validate the resulting facts by leveraging global constraints inherent in large fact collections, observing that correct facts will tend to match their arguments with other facts more often than with incorrect ones. We model this intuition as a graph-ranking problem over a fact graph and explore novel random walk algorithms. We present an empirical study, over a large set of facts extracted from a 500 million document webcrawl, validating the model and showing that it improves fact quality over state-of-the-art methods.