Medical Term and Status Generation From Chinese Clinical Dialogue With Multi-Granularity Transformer

This paper describes a generative model for extracting medical terms and their status from Chinese medical dialogues. Notably, the extracted semantic information is particularly important to downstream tasks like automatic medical scribe and automatic diagnosis systems. However, how to effectively leverage dialogue context to generate medical terms and their corresponding status accurately remains less explored. Existing generative approaches treat dialogue text as a single continuous text, ignoring conversational characteristics like colloquialism, redundancy and interactions. Between the doctor and the patient, a variety of colloquial medical information is frequently discussed. Each speaker (doctor and patient) plays a specific role in the interaction's goals. As a result, the importance of role information and interactions between utterances cannot be overstated. Furthermore, existing generative approaches only use character-level tokens, disregarding word-level tokens, which are the shortest meaningful utterances in Chinese. In this paper, we propose a Multi-granularity Transformer (MGT) model to enhance the dialogue context understanding from multi-granularity features. We incorporate word-level information by adapting a Lattice-based encoder with our proposed relative position encoding method. We further propose a Role Access Controlled Attention (RaCa) mechanism for introducing utterance-level interaction information. Experimental results on two benchmark datasets illustrate our model's validity and effectiveness, achieving state-of-the-art performance on both datasets.