A BERT-Span Model for Named Entity Recognition in Rehabilitation Medicine

Abstract Background.Due to multiple reasons such as the increasing aging of the population and the upgrading of people's health consumption needs, the demand group for rehabilitation medical care is expanding. At present, there are many problems in China's rehabilitation medical care, such as insufficient awareness and shortage of talents. It is especially important to enhance public awareness of rehabilitation and improve the quality of rehabilitation services. Named entity recognition, as the initial task of information processing, can automatically extract rehabilitation medical entities to serve downstream tasks such as information decision system and medical knowledge graph. Methods.To this end, we construct the BERT-Span model to complete the rehabilitation medicine named entity recognition task. First, we collect rehabilitation information from multiple sources to build a corpus in the field of rehabilitation medicine, and fine-tune BERT with the rehabilitation medicine corpus. For rehabilitation medicine text, we use BERT to extract the feature vectors of rehabilitation medicine entities in the text, and use the span model to complete the annotation of rehabilitation medicine entities. Result.Compared to existing baseline models, our model obtained the highest F1 value for the named entity recognition task in the rehabilitation medicine corpus.Conclusions.The experiment results show that our method achieves better results in both long medical entities and nested medical entities recognition in rehabilitation medical texts.