Neural Path Hunter: Reducing Hallucination in Dialogue Systems via Path Grounding

Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 2197–2214 November 7–11, 2021. c⃝2021 Association for Computational Linguistics 2197 Neural Path Hunter: Reducing Hallucination in Dialogue Systems via Path Grounding Nouha Dziri∗, Andrea Madotto†, Osmar Zaiane∗§, Avishek Joey Bose‡ ∗University of Alberta, ‡Mila, McGill University, §Canada CIFAR AI Chair †The Hong Kong University of Science and Technology dziri@cs.ualberta.ca Abstract Dialogue systems powered by large pre- trained language models exhibit an innate ability to deliver fluent and natural-sounding responses. Despite their impressive perfor- mance, these models are fitful and can often generate factually incorrect statements imped- ing their widespread adoption. In this paper, we focus on the task of improving faithful- ness and reducing hallucination of neural di- alogue systems to known facts supplied by a Knowledge Graph (KG). We propose NEU- RAL PATH HUNTER which follows a generate- then-refine strategy whereby a generated re- sponse is amended using the KG. NEURAL PATH HUNTER leverages a separate token- level fact critic to identify plausible sources of hallucination followed by a refinement stage that retrieves correct entities by crafting a query signal that is propagated over a k-hop subgraph. We empirically validate our pro- posed approach on the OpenDialKG dataset (Moon et al., 2019) against a suite of metrics and report a relative improvement of faithful- ness over dialogue responses by 20.35% based on FeQA (Durmus et al., 2020). The code is available at https://github.com/ nouhadziri/Neural-Path-Hunter. 1 Introduction Conversation within a dialogue can be thought of as an exchange of utterances between two speakers. Each utterance is not independent of one another but is instead grounded within a larger dialogue context known to both parties (Jurafsky and Mar- tin, 2018; Sordoni et al., 2015; Serban et al., 2016; Dziri et al., 2019). Indeed, if a response to an utter- ance fails to be faithful to some given knowledge— i.e. by producing false information—it is uninfor- mative and runs the risk of jeopardizing the entire enterprise of conversation. More precisely, this means that in addition to being fluent, topical, and ∗ Corresponding author. Figure 1: NEURAL PATH HUNTER overview. grammatical, utterances within a dialogue must also be factually correct. The faithfulness of responses is of principal im- portance when designing dialogue systems that are grounded using auxiliary knowledge such as Knowledge Graphs (KG). Despite maintaining plausible general linguistic capabilities, dialogue models are still unable to fully discern facts and may instead hallucinate factually invalid informa- tion. Moreover, empirical evidence for hallucina- tion in Language Models (LM) runs contrary to known studies that these large models are capable of recalling factual knowledge, e.g. entities and relations in a KG, (Roberts et al., 2020; Petroni et al., 2019). This suggests that this inherent lack of controllability may be remedied by leveraging external oracle knowledge. However, existing ap- proaches to knowledge grounding often suffer from a source-reference divergence problem whereby the reference contains additional factual information and simply training on the reference is insufficient to guarantee faithfulness (Wiseman et al., 2017; Parikh …

Content truncated to 500 words