Document Classification for COVID-19 Literature

TL;DR

This paper evaluates various multi-label document classification models on the LitCovid dataset, demonstrating that fine-tuned pre-trained language models, especially BioBERT, achieve high accuracy in classifying COVID-19 research papers, and analyzes their errors and limitations.

Contribution

It provides a comprehensive analysis of classification models on COVID-19 literature, highlighting the effectiveness of fine-tuned BioBERT and identifying key challenges for future improvements.

Findings

BioBERT achieves micro-F1 of around 86% and accuracy of 75%.

Pre-trained language models outperform baselines on LitCovid.

Errors often involve label correlation and focus issues.

Abstract

The global pandemic has made it more important than ever to quickly and accurately retrieve relevant scientific literature for effective consumption by researchers in a wide range of fields. We provide an analysis of several multi-label document classification models on the LitCovid dataset, a growing collection of 23,000 research papers regarding the novel 2019 coronavirus. We find that pre-trained language models fine-tuned on this dataset outperform all other baselines and that BioBERT surpasses the others by a small margin with micro-F1 and accuracy scores of around 86% and 75% respectively on the test set. We evaluate the data efficiency and generalizability of these models as essential features of any system prepared to deal with an urgent situation like the current health crisis. Finally, we explore 50 errors made by the best performing models on LitCovid documents and find that they often (1) correlate certain labels too closely together and (2) fail to focus on discriminative sections of the articles; both of which are important issues to address in future work. Both data and code are available on GitHub.