ChiMST: A Chinese Medical Corpus for Word Segmentation and Medical Term Recognition

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 5654–5664 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 5654 ChiMST: A Chinese Medical Corpus for Word Segmentation and Medical Term Recognition Yuanhe Tian♥, Han Qin♠, Fei Xia♥, Yan Song♠† ♥University of Washington ♠The Chinese University of Hong Kong (Shenzhen) ♥{yhtian, fxia}@uw.edu ♠hanqin@link.cuhk.edu.cn ♠songyan@cuhk.edu.cn Abstract Chinese word segmentation (CWS) and named entity recognition (NER) are two important tasks in Chinese natural language processing. To achieve good model performance on these tasks, existing neural approaches normally require a large amount of labeled training data, which is often unavailable for specific domains such as the Chinese medical domain due to privacy and legal issues. To address this problem, we have developed a Chinese medical corpus named ChiMST which consists of question-answer pairs collected from an online medical healthcare platform and is annotated with word boundary and medical term information. For word boundary, we mainly follow the word segmentation guidelines for the Penn Chinese Treebank (Xia, 2000); for medical terms, we define 9 categories and 18 sub-categories after consulting medical experts. To provide baselines on this corpus, we train existing state-of-the-art models on it and achieve good performance. We believe that the corpus and the baseline systems will be a valuable resource for CWS and NER research on the medical domain. Keywords: Chinese medical corpus, Chinese word segmentation, named entity recognition 1. Introduction In the past decade, there have been tremendous progress in applying natural language processing (NLP) techniques to medical text, with the hope that the progress will lead to real-world applications that improve healthcare quality. When analyzing Chinese medical text, Chinese word segmentation (CWS) and named entity recognition (NER) are often two of the early steps whose outputs are needed by downstream modules, and they have thus attracted attentions from both academia and industry (e.g., (Xing et al., 2018; Wang et al., 2018; Zhang et al., 2019; Wang et al., 2019; Luo et al., 2019; Chang et al., 2021; Li et al., 2021)). Building high-quality systems for the two tasks often requires large labeled datasets, which are usually un- available for special domains such as the Chinese med- ical domain. While there are studies (Xu et al., 2014; Jinfeng et al., 2016; He et al., 2017; Gao et al., 2019; Xiong et al., 2019; Su et al., 2019; Zan et al., 2020; Zhang et al., 2020a) that annotated Chinese medical text with CWS and NER labels, most of them use elec- tronic medical records (EMR) collected from hospitals and thus are not publicly available due to privacy and legal issues. Meanwhile, little attention has been paid to medical text from online healthcare platforms espe- cially the ones with questions from patients and an- swers from doctors. In addition, the named entities in most studies use common categories such as disease and symptom, without including fine-grained medical term types such as abnormality types (e.g., 二型 (Type II)) and body functions (e.g., 消化 (digestion)). In …

Content truncated to 500 words