Cross-type Biomedical Named Entity Recognition with Deep Multi-Task Learning

Motivation State-of-the-art biomedical named entity recognition systems often require specific handcrafted features for each entity type. The feature generation process is time and labor consuming, and leads to highly specialized systems not adaptable to new entity types. Although recent studies explored using neural network models for BioNER to free experts from manual feature generation, the performance remains limited by the available training data for each entity type. Results We propose a multi-task learning framework for BioNER to collectively use the training data of different entity types and improve the performance on each of them. In experiments on five BioNER datasets covering four major biomedical entity types, our multi-task model outperforms state-of-the-art systems and other neural network models by a large margin. Further analysis shows that the large performance gains come from sharing character- and word-level information across different biomedical entities. Availability The source code for our models is available at https://github.com/yuzhimanhua/lm-lstm-crf, and the corpora are available at https://github.com/cambridgeltl/MTL-Bioinformatics-2016. Contact xwang174@illinois.edu, xiangren@usc.edu. Supplementary information Supplementary data are available at Bioinformatics online.