Abstract
Medical Dialogue Generation serves a critical role in telemedicine byfacilitating the dissemination of medical expertise to patients. Existingstudies focus on incorporating textual representations, which have limitedtheir ability to represent the semantics of text, such as ignoring importantmedical entities. To enhance the model's understanding of the textual semanticsand the medical knowledge including entities and relations, we introduce theuse of Abstract Meaning Representations (AMR) to construct graphicalrepresentations that delineate the roles of language constituents and medicalentities within the dialogues. In this paper, We propose a novel framework thatmodels dialogues between patients and healthcare professionals using AMRgraphs, where the neural networks incorporate textual and graphical knowledgewith a dual attention mechanism. Experimental results show that our frameworkoutperforms strong baseline models in medical dialogue generation,demonstrating the effectiveness of AMR graphs in enhancing the representationsof medical knowledge and logical relationships. Furthermore, to support futureresearch in this domain, we provide the corresponding source code athttps://github.com/Bernard-Yang/MedDiaAMR.