Deep Scaled Dot-Product Attention Based Domain Adaptation Model For Biomedical Question Answering.

•A hierarchical attention based transfer learning model is proposed for biomedical QA.•BERT is applied to enrich the semantic representation in our model.•Scaled dot-product attention captures the question interaction for passage encoding.•Fine-tuning is used to penalize the deviations from source model’s parameters.•The results show that our model outperforms the best solution on BioASQ Task B.