Reducing the Frequency of Hallucinated Quantities in Abstractive Summaries

By Zheng Zhao, Shay B. Cohen, Bonnie Lynn Webber. 2020

42 citations recorded by the source corpus.

It is well-known that abstractive summaries are subject to hallucination—including material that is not supported by the original text. While summaries can be made hallucination-free by limiting them to general phrases, such summaries would fail to be very informative. Alternatively, one can try to avoid hallucinations by verifying that any specific entities in the summary appear in the original text in a similar context. This is the approach taken by our system, Herman. The system learns to recognize and verify quantity entities (dates, numbers, sums of money, etc.) in a beam-worth of abstractive summaries produced by state-of-the-art models, in order to up-rank those summaries whose quantity terms are supported by the original text. Experimental results demonstrate that the ROUGE scores of such up-ranked summaries have a higher Precision than summaries that have not been up-ranked, without a comparable loss in Recall, resulting in higher F1. Preliminary human evaluation of up-ranked vs. original summaries shows people’s preference for the former.

Subfields: Artificial Intelligence, Natural Language Processing, Summarization

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 Natural Language Processing

In Summarization

Citations in the corpus

Language Generation Models Can Cause Harm: So What Can We Do About It? An Actionable Survey (2022, background): “, 2012; Burnap and Williams, 2016) Misinformation Word-Level features (Zhao et al., 2020; King et al., 2022)”
Survey of Hallucination in Natural Language Generation (2022, background): “However, the correctness of the numerals in the generated texts such as date, quantity and scalar are important for readers [168, 212, 216].”
Are Factuality Checkers Reliable? Adversarial Meta-evaluation of Factuality in Summarization (2021, background): “…automated metrics w.r.t factuality as different problems: text entailment over sequential (Kryscinski et al., 2020; Goyal and Durrett, 2021a) or tree (Goyal and Durrett, 2020, 2021a) structures; question answering (Wang et al., 2020a; Durmus et al., 2020) and sequence labeling (Zhao et al., 2020a).”
Increasing Faithfulness in Knowledge-Grounded Dialogue with Controllable Features (2021, background): “Many of these works focus on other generation tasks such as summarization (Maynez et al., 2020; Zhao et al., 2020; Cao et al., 2018; Falke et al., 2019) or data-to-text generation (Puduppully et al., 2019).”
Hallucinated but Factual! Inspecting the Factuality of Hallucinations in Abstractive Summarization (2021, methodology): “They investigate the 043 cause of model hallucination (Kang and Hashimoto, 044 2020; Wang and Sennrich, 2020) and propose meth- 045 ods that reduce the frequency of all hallucinations 046 (Filippova, 2020; Zhao et al., 2020; Nan et al., 047 2021; Narayan et al., 2021).”
Annotating and Modeling Fine-grained Factuality in Summarization (2021, methodology): “Localizing errors potentially allows for post-hoc correction (Zhao et al., 2020; Cao et al., 2020); however, repairing a summary to be fully factual is a very hard problem and past work has focused on a subset of errors as a result.”
How Optimal is Greedy Decoding for Extractive Question Answering? (2021, background): “In contrast with hallucination, commonly reported in summarization tasks (Lewis et al., 2020; Zhao et al., 2020), the information added in these answers is actually correct.”