Our goal is to develop an automated solution which accomplishes three tasks: 1) When given a document, it differentiates which sentences are making substantive claims that require verification and which do not; 2) When given sentences, it detects whether it agrees or disagrees with some evidence (for example, given the evidence The bear scared off the baby that was attacking him, does it agree or disagree with the claim The bear is scary ); and 3) When given a claim, it provides evidence from a large corpus, such as Wikipedia, that agrees or disagrees with that claim (for example, given the claim The bear is scary, it should be able to provide documents that prove or disprove this claim).

“I’m participating in SuperUROP because it’ll help provide a great venue to do some substantial research at MIT. I feel sufficiently prepared by a combination of classes and side projects. I hope to learn a lot about the state of natural language processing research; particularly, its applications in industry. Fake news’ is also something that really excites me, and I couldn’t be more grateful to work on it.”