DiaKoP: Dialogue-based Knowledge-oriented Programming for Neural-symbolic Knowledge Base Question Answering

Zhicheng Lee, Zhidian Huang, Zijun Yao, Jinxin Liu, Amy Xin, Lei Hou, Juanzi Li

Abstract: We present Dialogue-based Knowledge-oriented Programming system (DiaKoP), a system with a chat interface designed for multi-turn knowledge base question answering (KBQA). DiaKoP enables users to decompose complex questions into multiple simpler follow-up questions and interact with the system to obtain answers. Multi-turn KBQA presents unique challenges because users may switch topics or ask incomplete questions that rely on previous interactions. To address this, we develop a Dialogue History Tracker and Dialogue Policy to manage user conversations effectively. Additionally, we enhance the knowledge from the knowledge graph by integrating parametric knowledge from a large language model (LLM) to provide more comprehensive answers. To mitigate the issue of wrongly parsed questions by semantic parser, we implement a human-in-the-loop mechanism, allowing users to correct errors. We evaluate DiaKoP both qualitatively and quantitatively, with user study indicating that our system better meets users' needs. DiaKoP is open-sourced on https://github.com/THU-KEG/DiaKoP with a guiding demo on https://youtu.be/Tq17k0OxPVg.