用知识图谱提升对话生成的事实性，减少幻觉。
Improving Factuality for Dialogue Response Generation via Graph-Based Knowledge Augmentation
- 构建图文结合的知识图谱，引导对话重述与知识选择。
- 在两个数据集上事实得分提升超3%，优于当前最佳方法。
- 适合需要高准确性的对话系统开发者使用。
大型语言模型在自然语言处理任务中表现优异，但其容易产生幻觉——生成看似合理却与事实不符的内容，在对话回复生成中尤为突出。为缓解此问题，本文提出两种新型基于图结构的知识增强框架：文本化图谱对话回复生成（TG-DRG）和图感知对话回复生成（GA-DRG），通过推理引导的对话重述、对话语义知识选择与图增强的回复生成，提升回复的事实性。为评估回复事实性，提出一种对话事实得分（dialogue fact score），克服现有方法在对话场景中的局限性，提供更可靠的评估。在OpendialKG和HybriDialogue数据集上，与多种基线方法对比，本方法显著提升事实性，相较当前最优的G-retriever，在OpendialKG上提升3.47%，在HybriDialogue上提升3.12%。代码将开源于GitHub。
原文摘要 · Abstract (English)
Large Language Models (LLMs) succeed in many natural language processing tasks. However, their tendency to hallucinate - generate plausible but inconsistent or factually incorrect text - can cause significant problems in certain tasks, including response generation in dialogue. To mitigate this issue, we propose two novel graph knowledge-augmented frameworks, Dialogue Response Generation via Textualised Graphs (TG-DRG) and Graph-Aware Dialogue Response Generation (GA-DRG), which combine reasoning-guided dialogue reformulation, dialogue sense knowledge selection, and graph-enhanced response generation to improve the factuality of dialogue responses. To evaluate the factuality of generated responses, we propose a dialogue fact score that addresses the limitations of existing fact-score methods in dialogue settings, providing a more reliable assessment of factual consistency. We evaluate our methods using different baselines on the OpendialKG and HybriDialogue datasets. Our methods noticeably improve factuality compared to other graph knowledge-augmentation baselines, including the state-of-the-art G-retriever, achieving improvements of 3.47% on OpendialKG and 3.12% on HybriDialogue in terms of dialogue fact score. The code will be released on GitHub.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。