A Paragraph-level Multi-task Learning Model for Scientific Fact-Verification

TL;DR

This paper introduces a paragraph-level multi-task learning model using BERT for scientific fact verification, jointly training on rationale selection and stance prediction to combat misinformation.

Contribution

It presents a novel multi-task learning approach that directly computes contextualized sentence embeddings for scientific claim verification.

Findings

Improved accuracy in scientific fact verification tasks

Effective joint training on rationale selection and stance prediction

Enhanced handling of scientific claims and evidence

Abstract

Even for domain experts, it is a non-trivial task to verify a scientific claim by providing supporting or refuting evidence rationales. The situation worsens as misinformation is proliferated on social media or news websites, manually or programmatically, at every moment. As a result, an automatic fact-verification tool becomes crucial for combating the spread of misinformation. In this work, we propose a novel, paragraph-level, multi-task learning model for the SciFact task by directly computing a sequence of contextualized sentence embeddings from a BERT model and jointly training the model on rationale selection and stance prediction.