Abstract

We approach the BioCreative VI Track 1 task of biological entity identification by focusing on named entity recognition (NER) and linking tagged entities to standard database identifiers. For this task, we apply recent neural NER techniques of combining bi-directional long short term memory (BLSTM) network layers with conditional random fields (CRFs) to the biomedical domain. We then use context words, dictionary lookups, and external biological knowledge bases to match tagged biological entities with corresponding identifiers. Our system predicts cell types and cell lines, cellular components, organisms and species, proteins and genes, small molecules, and tissues and organs.

Date

October 9, 2025

Authors

Emily Sheng, Scott Miller, JS Ambite, Prem Natarajan

Journal

Proceedings of the BioCreative VI Workshop

Pages

24-27

Publisher

Bethesda, MD, USA