BioRED task DUTIR-901 submission: Enhancing Biomedical Document-Level Relation Extraction through Multi-Task Method

Abstract

Biomedical relation extraction is central to biomedical natural language processing and is crucial for various downstream applications. The BioCreative VIII BioRED Track focuses on extracting entity relationships from biomedical literature titles and abstracts and classifying relations that are novel findings. This paper describes our method used to create submissions to identify all the relationships between human-annotated entities and implement an end-to-end system to identify all the asserted relationship subtasks. In our method, a multi-task training approach is employed for fine-tuning a pre-trained language model in the field of biology. Based on a broad spectrum of carefully designed tasks, our multi-task method not only extracts relations of better quality due to more effective supervision, but also achieves a more accurate classification of whether the entity pairs are novel findings. The official results on the test set show that our best submission achieves the F1-scores of 0.4441 on Subtask1 and 0.2334 on Subtask2.

This article is part of the Proceedings of the BioCreative VIII Challenge and Workshop: Curation and Evaluation in the era of Generative Models.