BioNER

BioNER is a software tool for biomedical named entity recognition, which is a crucial task in biomedical literature mining. The tool utilizes a novel multi-task learning model with a cross-sharing structure to improve the performance of BioNER tasks. The key features and findings of BioNER are:

1. The cross-sharing structure in the multi-task model allows for utilizing features from multiple datasets during the training process, enhancing the model's performance.

2. Experiments demonstrate that BioNER outperforms other multi-task models on gene, protein, and disease categories datasets.

3. The tool explores the best dataset pairs for multi-task learning and analyzes the influence of different entity types using sub-datasets.

4. BioNER maintains positive results even when the dataset size is reduced, indicating its robustness.

5. BioNER's detailed analysis offers guidance on selecting appropriate dataset pairs for multi-task training in biomedical named entity recognition tasks.