Structured Named Entity Recognition (NER) in Biomedical Texts Using Pre-Trained Language Models

Abstract

The field of Natural Language Processing (NLP) has witnessed remarkable progress in recent years, particularly in the domain of biomedical text analysis. Named Entity Recognition (NER), a pivotal task in information extraction, holds the key to deciphering and extracting structured information from unstructured biomedical texts. Accurate identification and classification of entities, such as DNA, proteins, cell types, cell lines, and RNA, are imperative for advancing our comprehension of complex biological systems. This paper presents a comprehensive exploration of the application of state-of-the-art pre-trained language models, including Bert-base-cased, Distilbert-base-cased, Albert-base-V2, Xml-roberta-base, Ernie-2.0-base-en, and Conv-bert-base, for structured Named Entity Recognition in biomedical texts. The BioNLP2004 dataset, enriched with diverse entity types, forms the basis for our experiments. Our objectives encompass a thorough investigation into the effectiveness of different pre-trained language models for biomedical NER, an in-depth analysis of the challenges posed by the BioNLP2004 dataset, and a comparative evaluation of the selected models in terms of precision, recall, and F1 score. Additionally, we explore the impact of fine-tuning strategies on model performance. The insights gained from this research have the potential to advance the capabilities of language models in the biomedical domain, contributing to more efficient and precise biomedical text analysis. This work serves as a stepping stone towards the broader goals of bioinformatics and medical research. The paper concludes with a summary of findings and outlines potential avenues for future research. © 2024 IEEE.