Assessing The Factual Accuracy of Generated Text

By Ben Goodrich, Vinay Rao, Mohammad Saleh, Peter J. Liu. 2019

219 citations recorded by the source corpus.

We propose a model-based metric to estimate the factual accuracy of generated text that is complementary to typical scoring schemes like ROUGE (Recall-Oriented Understudy for Gisting Evaluation) and BLEU (Bilingual Evaluation Understudy). We introduce and release a new large-scale dataset based on Wikipedia and Wikidata to train relation classifiers and end-to-end fact extraction models. The end-to-end models are shown to be able to extract complete sets of facts from datasets with full pages of text. We then analyse multiple models that estimate factual accuracy on a Wikipedia text summarization task, and show their efficacy compared to ROUGE and other model-free variants by conducting a human evaluation study.

Subfields: Artificial Intelligence, Communication, Natural Language Processing, Evaluation & Benchmarks

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 Communication

In Natural Language Processing

In Evaluation & Benchmarks

Citations in the corpus

Reasoning Over Semantic-Level Graph for Fact Checking (2019, background): “There might be many different ways to construct the graph, such as open information extraction (Banko et al., 2007), named entity recognition plus relation classiﬁcation, sequence-to-sequence generation which is trained to produce structured tuples (Goodrich et al., 2019), etc.”
On Faithfulness and Factuality in Abstractive Summarization (2020, methodology): “Goodrich et al. (2019) proposed a modelbased automatic metric to assess the faithfulness of Wikipedia summaries; they trained an end-to-end model to extract a complete set of OpenIE-style (Banko et al., 2007) facts from both the source text and the generated summary.”
Survey of Hallucination in Natural Language Generation (2022, background): “[55] propose a fact extraction model to extract relation tuples (subject, relation, object).”
Evaluating the Factual Consistency of Abstractive Text Summarization (2019, methodology): “Experiments were conducted on the Gigaword (Graff and Cieri, 2003) dataset, through human evaluation the authors showed that the proposed technique substantially lowered the number of errors in generated single-sentence summaries.”
Asking and Answering Questions to Evaluate the Factual Consistency of Summaries (2020, methodology): “Goodrich et al. (2019) use information extraction models to measure factual overlap, but facts are restricted to pre-deﬁned schemas.”
FEQA: A Question Answering Evaluation Framework for Faithfulness Assessment in Abstractive Summarization (2020, methodology): “In addition to QA, recent work has used relation extraction and textual entailment models for faithfulness evaluation (Falke et al., 2019a; Goodrich et al., 2019).”
Q^{2}: Evaluating Factual Consistency in Knowledge-Grounded Dialogues via Question Generation and Question Answering (2021, methodology): “Evaluation approaches that address this gap were recently proposed for tasks like ∗ Work done during an internship at Google Research. machine translation and abstractive summarization (Sellam et al., 2020; Xu et al., 2020; Goodrich et al., 2019).”
Understanding Factuality in Abstractive Summarization with FRANK: A Benchmark for Factuality Metrics (2021, methodology): “Entailment Classification Goodrich et al. (2019); Kryscinski et al. (2020); Maynez et al. (2020); Goyal and Durrett (2020) model factuality as entailment classification breaking down the summary into smaller units, such as sentences, which are verified against the original article.”
Factual Error Correction for Abstractive Summarization Models (2020, methodology): “es have been proposed to detect or ensure the factual consistency of generated summaries, including using fact extraction or applying attention on fact triples (Cao et al., 2018; Zhang et al., 2019b; Goodrich et al., 2019), applying natural language inference or question answering models for consist”
GO FIGURE: A Meta Evaluation of Factuality in Summarization (2020, background): “Recent efforts by NLP researchers have drawn attention to the issue of factual errors and hallucinations in the output of neural summarization models (Cao et al., 2018; Massarelli et al., 2019; Zhao et al., 2020; Falke et al., 2019b; Goodrich et al., 2019; Celikyilmaz et al., 2020).”
Evaluating Factuality in Generation with Dependency-level Entailment (2020, background): “The ﬁrst group leverages structured knowledge, like Open IE triples (Cao et al., 2018; Goodrich et al., 2019), dependency trees (Song et al., 2018), or generated semantic roles (Fan et al., 2018) as additional input for generation.”
Factuality Enhanced Language Models for Open-Ended Text Generation (2022, background): “Hallucinated NE Error Since NEs are one of the core building blocks of “fact”, NE-related metric design is one of the common choices in literature [11, 63, 64].”
Logical Natural Language Generation from Open-Domain Tables (2020, methodology): “CO (Lin et al.,2014), IE-based extractive evaluation (Rohrbach et al., 2018;Dhingra et al.,2019) are adopted for surfacelevel matching to replace costly human evaluation. In abstractive summarization,Goodrich et al. (2019) proposes NER + Relation Classiﬁcation method to investigate ﬁdelity in genera”
Program Enhanced Fact Verification with Verbalization and Graph Attention Network (2020, background): “…has become crucial for many applications such as detecting fake news, ru-mors, and political deception (Rashkin et al., 2017; Thorne et al., 2018; Goodrich et al., 2019; Vaibhav et al., 2019; Kry ´ sci ´ nski et al., 2019 and processing evidences from unstructured text data (Liu et al., 2020;…”
A Comprehensive Survey of AI-Generated Content (AIGC): A History of Generative AI from GAN to ChatGPT (2023, methodology): “Specifically, [221] proposed a Transformer-based end-to-end fact extraction model, which enables the structured prediction of relation tuples for factualness assessment.”
A Token-level Reference-free Hallucination Detection Benchmark for Free-form Text Generation (2021, background): “, 2019), information extraction (Zhang et al., 2020; Goodrich et al., 2019) or question answering (Scialom et al.”
Evaluating Semantic Accuracy of Data-to-Text Generation with Natural Language Inference (2020, background): “Recently, Goodrich et al. (2019) Faithful NLG Some recent neural NLG systems train speciﬁcally for semantic accuracy (Nie et al., 2019; Tian et al., 2019; Kedzie and McKeown, 2019).”
Optimizing the Factual Correctness of a Summary: A Study of Summarizing Radiology Reports (2019, background): “As a result, even state-of-the-art systems trained with ample data still produce summaries with a substantial number of factual errors (Goodrich et al., 2019; Kry´sci´nski et al., 2019a).”
Enhancing Factual Consistency of Abstractive Summarization (2021, background): “Several approaches have been proposed to evaluate a summary’s factual consistency (Kry´sci´nski et al., 2019a; Goodrich et al., 2019; Maynez et al., 2020).”
Hallucination is Inevitable: An Innate Limitation of Large Language Models (2024, background): “For example, factual-centred metrics [19, 20, 40, 57] and benchmarks [34, 35, 65] have been proposed to measure and reduce hallucination on specific datasets.”