A Dataset for N-ary Relation Extraction of Drug Combinations

TL;DR

This paper introduces a novel expert-annotated dataset for extracting information about multi-drug combination efficacy from scientific literature, addressing a complex NLP challenge involving variable-length and cross-sentence relations.

Contribution

The paper presents the first dataset for n-ary relation extraction of drug combinations, highlighting its unique challenges and providing baseline models for future research.

Findings

Dataset enables extraction of multi-drug efficacy relations

Relations often span multiple sentences, requiring advanced language understanding

Baseline models show promising results with room for improvement

Abstract

Combination therapies have become the standard of care for diseases such as cancer, tuberculosis, malaria and HIV. However, the combinatorial set of available multi-drug treatments creates a challenge in identifying effective combination therapies available in a situation. To assist medical professionals in identifying beneficial drug-combinations, we construct an expert-annotated dataset for extracting information about the efficacy of drug combinations from the scientific literature. Beyond its practical utility, the dataset also presents a unique NLP challenge, as the first relation extraction dataset consisting of variable-length relations. Furthermore, the relations in this dataset predominantly require language understanding beyond the sentence level, adding to the challenge of this task. We provide a promising baseline model and identify clear areas for further improvement. We release our dataset, code, and baseline models publicly to encourage the NLP community to participate in this task.