Sticking to the Facts: Confident Decoding for Faithful Data-to-Text Generation

Tian, Rong; Narayan, Shashi; Sellam, Thibault; Parikh, Ankur P.

doi:10.48550/arxiv.1910.08684

Cited by 28 publications

(41 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Hallucination. The injection of false information is a well-known phenomena in data-to-text generation (Wiseman et al, 2017;Tian et al, 2019;Dhingra et al, 2019;Parikh et al, 2020), and machine translation (Koehn and Knowles, 2017;, image captioning (Rohrbach et al, 2018), exposure bias (Wang and Sennrich, 2020), and question answering (Feng et al, 2018). More related to the task of dialogue systems, Dušek et al (2018Dušek et al ( , 2020 demonstrate that state-of-the-art natural language generation (NLG) models suffer from the hallucination problem.…”

Section: Related Workmentioning

confidence: 99%

“…This suggests that this inherent lack of controllability may be remedied by leveraging external oracle knowledge. However, existing approaches 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 to the source (Wiseman et al, 2017;Parikh et al, 2020;Tian et al, 2019). Consequently, ensuring the faithfulness of knowledge grounded dialogue systems-via precise alignment of the source and reference-remains an open challenge.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

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

Dziri¹,

Madotto²,

Zaı̈ane³

et al. 2021

Preprint

View full text Add to dashboard Cite

Dialogue systems powered by large pretrained language models (LM) exhibit an innate ability to deliver fluent and naturallooking responses. Despite their impressive generation performance, these models can often generate factually incorrect statements impeding their widespread adoption. In this paper, we focus on the task of improving the faithfulness-and thus reduce hallucinationof Neural Dialogue Systems to known facts supplied by a Knowledge Graph (KG). We propose NEURAL PATH HUNTER which follows a generate-then-refine strategy whereby a generated response is amended using the k-hop subgraph of a KG. NEURAL PATH HUNTER leverages a separate token-level fact critic to identify plausible sources of hallucination followed by a refinement stage consisting of a chain of two neural LM's that retrieves correct entities by crafting a query signal that is propagated over the k-hop subgraph. Our proposed model can easily be applied to any dialogue generated responses without retraining the model. We empirically validate our proposed approach on the OpenDialKG dataset (Moon et al., 2019) against a suite of metrics and report a relative improvement of faithfulness over GPT2 dialogue responses by 8.4%.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

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

Dziri¹,

Madotto²,

Zaı̈ane³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…To reduce hallucinations in the reference-based setting, researchers have applied iterative training (Nie et al, 2019), post editing (Dong et al, 2020), soft constraints, e.g. attention manipulation (Kiddon et al, 2016;Hua and Wang, 2019;Tian et al, 2019; or optimal transport (Wang et al, 2020b), and template/scaffold guided schema (Liu et al, 2017;Wiseman et al, 2018;Moryossef et al, 2019;Ye et al, 2020;Shen et al, 2020;Li and Rush, 2020;Balakrishnan et al, 2019;Liu et al, 2021).…”

Section: Related Workmentioning

confidence: 99%

A Token-level Reference-free Hallucination Detection Benchmark for Free-form Text Generation

Li¹,

Zhang²,

Brockett³

et al. 2021

Preprint

View full text Add to dashboard Cite

Large pretrained generative models like GPT-3 often suffer from hallucinating non-existent or incorrect content, which undermines their potential merits in real applications. Existing work usually attempts to detect these hallucinations based on a corresponding oracle reference at a sentence or document level. However ground-truth references may not be readily available for many free-form text generation applications, and sentence-or document-level detection may fail to provide the fine-grained signals that would prevent fallacious content in real time. As a first step to addressing these issues, we propose a novel token-level, reference-free hallucination detection task and an associated annotated dataset named HADES (HAllucination DEtection dataSet) 1 . To create this dataset, we first perturb a large number of text segments extracted from English language Wikipedia, and then verify these with crowd-sourced annotations. To mitigate label imbalance during annotation, we utilize an iterative model-inloop strategy. We conduct comprehensive data analyses and create multiple baseline models.

show abstract

“…Concept Expansion. With limited number of entities and concepts as input, generation systems are often incapable of producing long text with rich content, resulting in hallucination (Wiseman et al, 2017;Tian et al, 2019). Therefore, from the often-abstract core concepts, we aim to predict more specific concepts that are also relevant to the given title.…”

Section: Content Item Augmentationmentioning

confidence: 99%

DYPLOC: Dynamic Planning of Content Using Mixed Language Models for Text Generation

Hua

Sreevatsa²,

Wang

2021

Preprint

View full text Add to dashboard Cite

We study the task of long-form opinion text generation, which faces at least two distinct challenges. First, existing neural generation models fall short of coherence, thus requiring efficient content planning. Second, diverse types of information are needed to guide the generator to cover both subjective and objective content. To this end, we propose DY-PLOC, a generation framework that conducts dynamic planning of content while generating the output based on a novel design of mixed language models. To enrich the generation with diverse content, we further propose to use large pre-trained models to predict relevant concepts and to generate claims. We experiment with two challenging tasks on newly collected datasets: (1) argument generation with Reddit ChangeMyView, and (2) writing articles using New York Times' Opinion section. Automatic evaluation shows that our model significantly outperforms competitive comparisons. Human judges further confirm that our generations are more coherent with richer content.

show abstract

Sticking to the Facts: Confident Decoding for Faithful Data-to-Text Generation

Cited by 28 publications

References 26 publications

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

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

A Token-level Reference-free Hallucination Detection Benchmark for Free-form Text Generation

DYPLOC: Dynamic Planning of Content Using Mixed Language Models for Text Generation

Contact Info

Product

Resources

About