MTL-BC-LBC-BioNER

MTL-BC-LBC-BioNER implements dataset-aware multi-task learning to jointly train Bio-NER and biomedical part-of-speech (POS) tagging models across multiple Bio-NER datasets to improve entity recognition performance and mitigate data scarcity.

Key Features:

• Dataset-Aware Multi-Task Learning: Implements two dataset-aware MTL approaches that leverage diverse information across multiple Bio-NER datasets and enable models to discriminatively exploit related training datasets.
• Performance Enhancement: Joint training across numerous Bio-NER datasets outperforms existing state-of-the-art MTL methods on 14 of 15 Bio-NER datasets.
• Integration with Biomedical POS Tagging: Incorporates Bio-NER and biomedical part-of-speech (POS) tagging datasets to mutually enhance the performance of both tasks.

Scientific Applications:

• Biomedical Text Mining: Identifies and classifies biomedical entities in text to facilitate information extraction from literature in genomics, proteomics, and drug discovery.
• Data-Driven Research: Provides a framework that leverages information across datasets to handle limited high-quality datasets, supporting research under data-scarce conditions.

Methodology:

The methodology develops two dataset-aware multi-task learning approaches that jointly train models across multiple Bio-NER datasets and integrate biomedical POS tagging datasets.