Diving Deep into Modes of Fact Hallucinations in Dialogue Systems

TL;DR

This paper investigates different types of fact hallucinations in knowledge graph-grounded dialogue systems, introduces a synthetic dataset for hallucination detection, and evaluates baseline models to improve response accuracy.

Contribution

It systematically identifies hallucination modes, creates the FADE dataset using perturbation strategies, and benchmarks detection models for better control of factual errors.

Findings

Identified various hallucination modes through human feedback analysis.

Developed the FADE dataset for training and evaluating detection models.

Established baseline models for hallucination detection and compared with human annotations.

Abstract

Knowledge Graph(KG) grounded conversations often use large pre-trained models and usually suffer from fact hallucination. Frequently entities with no references in knowledge sources and conversation history are introduced into responses, thus hindering the flow of the conversation -- existing work attempt to overcome this issue by tweaking the training procedure or using a multi-step refining method. However, minimal effort is put into constructing an entity-level hallucination detection system, which would provide fine-grained signals that control fallacious content while generating responses. As a first step to address this issue, we dive deep to identify various modes of hallucination in KG-grounded chatbots through human feedback analysis. Secondly, we propose a series of perturbation strategies to create a synthetic dataset named FADE (FActual Dialogue Hallucination DEtection Dataset). Finally, we conduct comprehensive data analyses and create multiple baseline models for hallucination detection to compare against human-verified data and already established benchmarks.