CliCR: A Dataset of Clinical Case Reports for Machine Reading Comprehension

TL;DR

This paper introduces CliCR, a large dataset of clinical case reports designed for machine reading comprehension, highlighting the challenges and skills needed for effective medical question answering.

Contribution

It provides a new dataset with 100,000 clinical queries and analyzes the skills and difficulties faced by neural readers in medical comprehension tasks.

Findings

Significant performance gap between humans and machines (20% F1).

Domain knowledge and object tracking are key skills for successful answering.

Recognizing omitted info and spatio-temporal reasoning are most challenging for models.

Abstract

We present a new dataset for machine comprehension in the medical domain. Our dataset uses clinical case reports with around 100,000 gap-filling queries about these cases. We apply several baselines and state-of-the-art neural readers to the dataset, and observe a considerable gap in performance (20% F1) between the best human and machine readers. We analyze the skills required for successful answering and show how reader performance varies depending on the applicable skills. We find that inferences using domain knowledge and object tracking are the most frequently required skills, and that recognizing omitted information and spatio-temporal reasoning are the most difficult for the machines.