Figure 1
From: Knowledge grounded medical dialogue generation using augmented graphs

Medical Knowledge Graph: The patient-doctor dialogue as shown on the left is first passed through the Quick-UMLS tool, which is used to extract the medical concepts and relations from the UMLS knowledge base as shown in the box on top-middle. The box on the bottom-middle uses the MEP model to discover additional biomedical words present in the dialogues. Finally, the augmented knowledge graph used in the dialogue generation pipeline is shown in the right.