# Team Papelo: Transformer Networks at FEVER

## TL;DR

This paper presents a transformer-based system for FEVER fact verification that uses high-precision entailment classification to improve evidence recall and achieves competitive scores on the shared task.

## Contribution

It introduces a high-precision transformer entailment classifier that enhances evidence retrieval by considering multiple articles and contextual clues, improving FEVER scores.

## Key findings

- Achieved 0.5736 FEVER score on test set
- Improved evidence recall through multi-article analysis
- Utilized transformer-based entailment for better classification

## Abstract

We develop a system for the FEVER fact extraction and verification challenge that uses a high precision entailment classifier based on transformer networks pretrained with language modeling, to classify a broad set of potential evidence. The precision of the entailment classifier allows us to enhance recall by considering every statement from several articles to decide upon each claim. We include not only the articles best matching the claim text by TFIDF score, but read additional articles whose titles match named entities and capitalized expressions occurring in the claim text. The entailment module evaluates potential evidence one statement at a time, together with the title of the page the evidence came from (providing a hint about possible pronoun antecedents). In preliminary evaluation, the system achieves .5736 FEVER score, .6108 label accuracy, and .6485 evidence F1 on the FEVER shared task test set.