BioRED track lasigeBioTM submission: Relation Extraction using Domain Ontologies with BioRED

Abstract

Biomedical relation extraction is a crucial task for extracting valuable knowledge from unstructured scientific literature. This paper discusses our team's, lasigeBioTM, involvement in the BioCreative VIII Track 1: BioRED in both sub-tasks. Our primary focus was on the relation extraction (RE) task, taking advantage of the K-RET system in combination with Gene Ontology, Chemical Entities of Biological Interest, Human Phenotype Ontology, Human Disease Ontology and NCBITaxon Ontology. The objective was to evaluate whether the use of external knowledge could enhance the performance of the relation extraction task, both for entity relationships and for detecting novel information. Our results in both tasks were below the average and we were not able to discern the impact of the introduced external knowledge. However, it was observed that for our model, a cleaner dataset is needed for improved performance and the necessity for a larger number of example instances, as our model struggled to identify low-represented labels.

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