A Simple Recipe towards Reducing Hallucination in Neural Surface Realisation

By Feng Nie, Jin-ge Yao, Jinpeng Wang, Rong Pan, Chin-Yew Lin. 2019

90 citations recorded by the source corpus.

Recent neural language generation systems often hallucinate contents (i.e., producing irrelevant or contradicted facts), especially when trained on loosely corresponding pairs of the input structure and text. To mitigate this issue, we propose to integrate a language understanding module for data refinement with self-training iterations to effectively induce strong equivalence between the input data and the paired text. Experiments on the E2E challenge dataset show that our proposed framework can reduce more than 50% relative unaligned noise from the original data-text pairs. A vanilla sequence-to-sequence neural NLG model trained on the refined data has improved on content correctness compared with the current state-of-the-art ensemble generator.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Natural Language Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Natural Language Processing

Citations in the corpus

Semantic Noise Matters for Neural Natural Language Generation (2019, methodology): “Contemporaneous with our work is the effort of Nie et al. (2019), who focus on automatic data cleaning using a NLU iteratively bootstrapped from the noisy data.”
Structure-to-Text Generation with Self-Training, Acceptability Classifiers and Context-Conditioning for the GEM Shared Task (2021, methodology): “Most approaches to self-training for NLG— including earlier work on automatic data cleaning— make use of cycle consistency between parsing and generation models (Chisholm et al., 2017; Nie et al., 2019; Kedzie and McKeown, 2019; Qader et al., 2019).”
Survey of Hallucination in Natural Language Generation (2022, methodology): “Currently, some works [137, 180, 230] point out that using commonsense knowledge can help to gain better numeral representation.”
Get Your Vitamin C! Robust Fact Verification with Contrastive Evidence (2021, methodology): “The FEVER dataset (Thorne et al., 2018) fueled the development of many fact-checking models (e.g., see Hanselowski et al., 2018; Nie et al., 2019a,b; Yoneda et al., 2018, inter alia ).”
On Hallucination and Predictive Uncertainty in Conditional Language Generation (2021, unknown): “Nie et al. (2019) believes that the origin of the hallucination problem in neural surface realization comes from the data side.”
Controlled Hallucinations: Learning to Generate Faithfully from Noisy Data (2020, unknown): “The latter approach of enforcing a stronger alignment between inputs and outputs has been tried previously but it assumes a moderate amount of noise in the data (Nie et al., 2019; Duˇsek et al., 2019).”
Sometimes We Want Ungrammatical Translations (2021, background): “…input to an output that is not faithful to the source, and sometimes is even factually incorrect (Vinyals and Le, 2015; Koehn and Knowles, 2017; Wiseman et al., 2017; Nie et al., 2019; Kryscinski et al., 2020; Maynez et al., 2020; Tian et al., 2020; González et al., 2020; Xiao and Wang, 2021).”
Plausible May Not Be Faithful: Probing Object Hallucination in Vision-Language Pre-training (2022, methodology): “Nie et al. (2019) applied data refinement with self-training to improve the equivalence between the input and the paired text in the data-to-text generation task.”
AlignScore: Evaluating Factual Consistency with A Unified Alignment Function (2023, background): “For a fair comparison, we also report the average AUC-ROC (denoted as AVG-ZS) excluding datasets that the alignment function is trained on (PAWS, VitaminC and FEVER).”
Value-Agnostic Conversational Semantic Parsing (2021, background): “String values can only enter the program through copying, as they are not in the set of constants (i.e., they cannot be “hallucinated” by the model; see Pasupat and Liang, 2015; Nie et al., 2019).”
Hallucination is Inevitable: An Innate Limitation of Large Language Models (2024, methodology): “For data-related issues, solutions include fact-focused datasets [17, 21] and developing automatic data cleaning techniques [42, 51, 54].”
Decoder-Only or Encoder-Decoder? Interpreting Language Model as a Regularized Encoder-Decoder (2023, background): “Generation models prone to generate unrelated content beyond the input data and this phenomenon has been recognized as the hallucination problem (Tian et al., 2019; Nie et al., 2019; Liu et al., 2021; Rebuffel et al., 2021) in many recent research works.”
Evidence-based Factual Error Correction (2020, background): “Without explicit control (Nie et al., 2019), the most likely token when decoded may not be factually accurate, or supported by the retrieved evidence, commonly referred to as a hallucination (Rohrbach et al., 2018; Zhou et al., 2020).”
Slot-consistent NLG for Task-oriented Dialogue Systems with Iterative Rectification Network (2020, background): “However, this step can introduce missing and misplaced slots, because of modeling errors or unaligned training data (Balakrishnan et al., 2019; Nie et al., 2019; Juraska et al., 2018).”
Building Adaptive Acceptability Classifiers for Neural NLG (2021, background): “Also related are earlier approaches that use cycle consistency between parsing and generation models for automatic data cleaning (Nie et al., 2019; Chisholm et al., 2017).”
A Token-level Reference-free Hallucination Detection Benchmark for Free-form Text Generation (2021, methodology): “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; Sha et al., 2018; Hua and Wang, 2019; Tian et al., 2019; Liu et al., 2019)…”
AggGen: Ordering and Aggregating while Generating (2021, background): “The prevailing sequence-to-sequence (seq2seq) architectures typically address this issue via reranking (Wen et al., 2015a; Dušek and Jurčı́ček, 2016; Juraska et al., 2018) or some sophisticated training techniques (Nie et al., 2019; Kedzie and McKeown, 2019; Qader et al., 2019).”
Factuality Enhanced Language Models for Open-Ended Text Generation (2022, background): “Hallucination in downstream NLG tasks There are active efforts to reduce the unfaithfulness or factual errors of task-speciﬁc LMs ﬁne-tuned for various downstream natural language generation (NLG) tasks such as summarization [42–48], data-to-text [49, 50, 20, 51–53] and dialogue system [54–58].”