Hierarchical Evidence Set Modeling for Automated Fact Extraction and Verification

Automated fact extraction and verification is a challenging task that\ninvolves finding relevant evidence sentences from a reliable corpus to verify\nthe truthfulness of a claim. Existing models either (i) concatenate all the\nevidence sentences, leading to the inclusion of redundant and noisy\ninformation; or (ii) process each claim-evidence sentence pair separately and\naggregate all of them later, missing the early combination of related sentences\nfor more accurate claim verification. Unlike the prior works, in this paper, we\npropose Hierarchical Evidence Set Modeling (HESM), a framework to extract\nevidence sets (each of which may contain multiple evidence sentences), and\nverify a claim to be supported, refuted or not enough info, by encoding and\nattending the claim and evidence sets at different levels of hierarchy. Our\nexperimental results show that HESM outperforms 7 state-of-the-art methods for\nfact extraction and claim verification. Our source code is available at\nhttps://github.com/ShyamSubramanian/HESM.\n