Unsupervised Question Answering for Fact-Checking

TL;DR

This paper introduces an unsupervised question-answering approach using BERT for fact-checking, transforming the FEVER dataset into a Cloze-task to classify claims with over 80% accuracy.

Contribution

It presents a novel unsupervised method for fact-checking by transforming datasets into Cloze-tasks and leveraging pre-trained language models like BERT.

Findings

Achieved 80.2% accuracy on the development set

Achieved 80.25% accuracy on the test set

Demonstrated effectiveness of unsupervised QA for fact-checking

Abstract

Recent Deep Learning (DL) models have succeeded in achieving human-level accuracy on various natural language tasks such as question-answering, natural language inference (NLI), and textual entailment. These tasks not only require the contextual knowledge but also the reasoning abilities to be solved efficiently. In this paper, we propose an unsupervised question-answering based approach for a similar task, fact-checking. We transform the FEVER dataset into a Cloze-task by masking named entities provided in the claims. To predict the answer token, we utilize pre-trained Bidirectional Encoder Representations from Transformers (BERT). The classifier computes label based on the correctly answered questions and a threshold. Currently, the classifier is able to classify the claims as "SUPPORTS" and "MANUAL_REVIEW". This approach achieves a label accuracy of 80.2% on the development set and 80.25% on the test set of the transformed dataset.