Advancing entity recognition in biomedicine via instruction tuning of large language models.
Bioinformatics
; 40(4)2024 03 29.
Article
em En
| MEDLINE
| ID: mdl-38514400
ABSTRACT
MOTIVATION Large Language Models (LLMs) have the potential to revolutionize the field of Natural Language Processing, excelling not only in text generation and reasoning tasks but also in their ability for zero/few-shot learning, swiftly adapting to new tasks with minimal fine-tuning. LLMs have also demonstrated great promise in biomedical and healthcare applications. However, when it comes to Named Entity Recognition (NER), particularly within the biomedical domain, LLMs fall short of the effectiveness exhibited by fine-tuned domain-specific models. One key reason is that NER is typically conceptualized as a sequence labeling task, whereas LLMs are optimized for text generation and reasoning tasks. RESULTS:
We developed an instruction-based learning paradigm that transforms biomedical NER from a sequence labeling task into a generation task. This paradigm is end-to-end and streamlines the training and evaluation process by automatically repurposing pre-existing biomedical NER datasets. We further developed BioNER-LLaMA using the proposed paradigm with LLaMA-7B as the foundational LLM. We conducted extensive testing on BioNER-LLaMA across three widely recognized biomedical NER datasets, consisting of entities related to diseases, chemicals, and genes. The results revealed that BioNER-LLaMA consistently achieved higher F1-scores ranging from 5% to 30% compared to the few-shot learning capabilities of GPT-4 on datasets with different biomedical entities. We show that a general-domain LLM can match the performance of rigorously fine-tuned PubMedBERT models and PMC-LLaMA, biomedical-specific language model. Our findings underscore the potential of our proposed paradigm in developing general-domain LLMs that can rival SOTA performances in multi-task, multi-domain scenarios in biomedical and health applications. AVAILABILITY AND IMPLEMENTATION Datasets and other resources are available at https//github.com/BIDS-Xu-Lab/BioNER-LLaMA.
Texto completo:
1
Base de dados:
MEDLINE
Assunto principal:
Camelídeos Americanos
/
Aprendizado Profundo
Limite:
Animals
Idioma:
En
Revista:
Bioinformatics
Assunto da revista:
INFORMATICA MEDICA
Ano de publicação:
2024
Tipo de documento:
Article
País de afiliação:
Estados Unidos