End-to-End Knowledge-Routed Relational Dialogue System for Automatic Diagnosis

TLDR

Automatic medical diagnosis dialogue systems face challenges such as topic transition coherency and question understanding, and existing data‑driven systems cannot encode expert knowledge graphs. The study proposes an End‑to‑End Knowledge‑routed Relational Dialogue System that automatically diagnoses patients by conversing to gather additional symptoms beyond self‑reports, leveraging a medical knowledge graph. KR‑DS uses a Knowledge‑routed Deep Q‑network that combines a relational refinement branch for symptom–disease relations and a knowledge‑routed graph branch to guide topic transitions, integrating the medical knowledge graph into dialogue management. KR‑DS outperforms state‑of‑the‑art methods by over 8 % in diagnosis accuracy on a public dataset and remains superior on a newly collected, more challenging dataset that preserves original self‑reports and patient‑doctor conversations.

Abstract

Beyond current conversational chatbots or task-oriented dialogue systems that have attracted increasing attention, we move forward to develop a dialogue system for automatic medical diagnosis that converses with patients to collect additional symptoms beyond their self-reports and automatically makes a diagnosis. Besides the challenges for conversational dialogue systems (e.g. topic transition coherency and question understanding), automatic medical diagnosis further poses more critical requirements for the dialogue rationality in the context of medical knowledge and symptom-disease relations. Existing dialogue systems (Madotto, Wu, and Fung 2018; Wei et al. 2018; Li et al. 2017) mostly rely on datadriven learning and cannot be able to encode extra expert knowledge graph. In this work, we propose an End-to-End Knowledge-routed Relational Dialogue System (KR-DS) that seamlessly incorporates rich medical knowledge graph into the topic transition in dialogue management, and makes it cooperative with natural language understanding and natural language generation. A novel Knowledge-routed Deep Q-network (KR-DQN) is introduced to manage topic transitions, which integrates a relational refinement branch for encoding relations among different symptoms and symptomdisease pairs, and a knowledge-routed graph branch for topic decision-making. Extensive experiments on a public medical dialogue dataset show our KR-DS significantly beats stateof-the-art methods (by more than 8% in diagnosis accuracy). We further show the superiority of our KR-DS on a newly collected medical dialogue system dataset, which is more challenging retaining original self-reports and conversational data between patients and doctors.