Semi-Supervised Noisy Label Learning for Chinese Medical Named Entity Recognition

This paper describes our approach for the Chinese clinical named entity recognition (CNER) task organized by 2020 China Conference on Knowledge Graph and Semantic Computing (CCKS) competition. In this task, we need to identify boundary category labels of six entities from electronic medical record (EMR). We constructed a hybrid system composed semi-supervised noisy label learning model based adversarial training rule post-processing module. The core idea is reduce impact data noise optimizing results. Besides, used rules correct three cases redundant labeling, missing wrong labeling in prediction Our method proposed achieved strict criteria 0.9156 relax 0.9660 final test set, ranking first.