Biomedical named entity recognition with the combined feature attention and fully-shared multi-task learning.

Area of Science:

• Computational Biology
• Natural Language Processing
• Bioinformatics

Background:

• Biomedical named entity recognition (BioNER) is crucial for biomedical text mining.
• Deep learning, particularly pre-trained language models like BioBERT, has advanced BioNER.
• Limitations exist due to scarce annotated data and external knowledge.

Purpose of the Study:

• To propose a novel fully-shared multi-task learning model for BioNER.
• To enhance BioNER performance by integrating auto-processed syntactic information via an attention module.
• To leverage BioBERT for improved biomedical entity recognition.

Main Methods:

• Developed a fully-shared multi-task learning framework.
• Integrated a novel attention module to incorporate syntactic information.
• Utilized the BioBERT pre-trained language model.

Main Results:

• The proposed model achieved F1 score improvements on seven benchmark BioNER datasets.
• Specific improvements ranged from 0.81% to 1.26% compared to single-task BioBERT.
• The model demonstrated superior performance across all tested datasets.

Conclusions:

• The developed model outperforms existing studies in BioNER.
• The attention module and multi-task learning approach are key to the model's success.
• Further analysis validates the effectiveness of the proposed methods.