Abstract
AbstractChinese medical named entity recognition (NER) task usually lacks sufficient annotation data, and it contains many medical professional terms and abbreviations, making the NER task more difficult. In addition, compared with English NER, Chinese NER is more challenging because it lacks standard feature symbols to determine named entity boundaries. Therefore, Chinese NER needs to perform word segmentation. In this paper, we are inspired by lexicon-based BERT and propose a novel method for Chinese medical NER task. Besides, We design a template-based strategy to enrich the words’ information and improve the model’s ability to distinguish medical professional terms and abbreviations. Our method enhances the word segmentation accuracy by introducing the external medical lexicon. To verify the effectiveness of our method, we carry out experiments on three medical datasets and our method improves them by 0.92%, 1.18% and 1.55% F1-score compared to baseline.KeywordsChinese medical NERExternal knowledgePrompt
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call