Edit 2019/Cure My FEVER : Building, Breaking and Fixing Models for Fact-Checking: Submissions:2019/“Wikipedians are born, not made” – applying the learnings from a 10-year-old research paper after all

This submission has been accepted for WikiConference North America 2019.

Many tasks such as question answering and reading comprehension rely on information extracted from unreliable sources. These systems would thus benefit from knowing whether a statement from an unreliable source is correct. We present experiments on the Fact Extraction and VERification (FEVER) dataset, which involves selecting sentences from Wikipedia and predicting whether a claim is supported by those sentences, refuted, or there is not enough information. We describe results across three phases: the development of a fact-checking system, the creation of adversarial examples, and the improvement of the system in handling adversarial examples.