What is disputed on the web?

ABSTRACTWe present a method for automatically acquiring of a corpus of disputed claims from the web. We consider a factual claim to be disputed if a page on the web suggests both that the claim is false and also that other people say it is true. Our tool extracts disputed claims by searching the web for patterns such as "falsely claimed that X" and then using a statistical classifier to select text that appears to be making a disputed claim. We argue that such a corpus of disputed claims is useful for a wide range of applications related to information credibility on the web, and we report what our current corpus reveals about what is being disputed on the web.