Probing Pre-Trained Language Models for Disease Knowledge

TL;DR

This paper introduces DisKnE, a new benchmark for evaluating disease knowledge and reasoning in pre-trained language models, revealing their limited medical reasoning capabilities.

Contribution

The paper presents DisKnE, a novel benchmark with adversarial negative examples and disease-specific splits to better assess medical reasoning in language models.

Findings

Pre-trained models perform poorly on the new benchmark.

Standard benchmarks may overestimate models' reasoning abilities.

DisKnE highlights the need for improved medical reasoning evaluation.

Abstract

Pre-trained language models such as ClinicalBERT have achieved impressive results on tasks such as medical Natural Language Inference. At first glance, this may suggest that these models are able to perform medical reasoning tasks, such as mapping symptoms to diseases. However, we find that standard benchmarks such as MedNLI contain relatively few examples that require such forms of reasoning. To better understand the medical reasoning capabilities of existing language models, in this paper we introduce DisKnE, a new benchmark for Disease Knowledge Evaluation. To construct this benchmark, we annotated each positive MedNLI example with the types of medical reasoning that are needed. We then created negative examples by corrupting these positive examples in an adversarial way. Furthermore, we define training-test splits per disease, ensuring that no knowledge about test diseases can be learned from the training data, and we canonicalize the formulation of the hypotheses to avoid the presence of artefacts. This leads to a number of binary classification problems, one for each type of reasoning and each disease. When analysing pre-trained models for the clinical/biomedical domain on the proposed benchmark, we find that their performance drops considerably.