Elamur Research⌕
☰
home / papers / 2021 / Q²: Evaluating Factual Consistency in Knowledge-Grounded Dialogues via Question Generation and Question Answering

Q²: Evaluating Factual Consistency in Knowledge-Grounded Dialogues via Question Generation and Question Answering

Summary

Knowledge-grounded dialogue systems often produce fluent replies that contradict or invent facts beyond the knowledge they were given. Standard metrics rarely catch this, and many need reference responses, which suit open-ended dialogue poorly. The authors propose Q², a reference-free metric that checks factual consistency in three steps:

  1. It extracts informative spans (named entities and noun phrases) from a system response.
  2. It generates questions whose answers are those spans.
  3. It answers the same questions against the grounding knowledge and compares the two answers.

Earlier QG/QA metrics for summarization compared answers by token overlap. Q² instead uses an NLI model, so paraphrases such as "LA" and "Los Angeles" still count as matches. It also filters out questions about personal opinions, which matters because dialogue mixes facts with chit-chat.

To test the metric, the authors built a new human-annotated dataset of consistent and inconsistent outputs from two systems on Wizard of Wikipedia. Across this dataset, Topical-Chat and DNLI, Q² correlates better with human judgements than competing reference-free metrics. This makes it a practical tool for assessing hallucination in grounded dialogue.

Key points

  • Pipeline:
    • spaCy selects the answer spans.
    • T5-base fine-tuned on SQuAD1.1 generates the questions (beam search, top-5 candidates).
    • Albert-Xlarge fine-tuned on SQuAD2.0 answers them and can return "no answer."
    • RoBERTa fine-tuned on SNLI (AllenNLP) compares the answers.
  • Scoring:
    • Entailment counts as 1, while contradiction or a missing answer counts as 0.
    • Neutral cases fall back to token-level F1.
    • If no valid question survives filtering, an end-to-end NLI prediction decides the score.
  • New dataset: responses from dodecaDialogue and MemNet on Wizard of Wikipedia, labeled consistent or inconsistent, in splits of 150 responses each. It is released on GitHub (orhonovich/q-squared).
  • Score separation: on dodecaDialogue, Q² gives about 0.24 to inconsistent responses versus about 0.70 to consistent ones. On MemNet, the figures are about 0.14 versus 0.76.
  • Sanity check with random knowledge: when the knowledge is swapped for a random passage, over 91% of questions find no answer. The rate rises to 99.6% when the passage comes from a different dialogue.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

Fields: Natural language processing

Cited by

83 papers of our library cite it (of the 140884 whose bibliographies we read); 11 of them have a page here, 72 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2021.emnlp-main.619

@inproceedings{honovich2021evaluating,
  title = {{Q²: Evaluating Factual Consistency in Knowledge-Grounded Dialogues via Question Generation and Question Answering}},
  author = {Honovich, Or and Choshen, Leshem and Aharoni, Roee and Neeman, Ella and Szpektor, Idan and Abend, Omri},
  booktitle = {Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing},
  year = {2021},
  doi = {10.18653/v1/2021.emnlp-main.619},
  url = {https://doi.org/10.18653/v1/2021.emnlp-main.619}
}

References

57 references, 17 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Automatic Evaluation of Dialogue Systems. Automatically evaluating natural language generation is a notoriously difficult problem, especially when considering open-ended tasks such as dialogue. Standard token-matching metrics, such as BLEU (Papineni et al., 2002) or METEOR (Banerjee and Lavie, 2005)… find
  2. Recently, Mehri and Eskenazi (2020) and Pang et al. (2020) suggested to use large pretrained language models (Liu et al., 2019; Radford et al., 2019) to develop reference-response-free metrics for dialogue evaluation. Such LMs are also the backbone of the QG, QA and NLI models employed in Q 2 . find
  3. Qualitative Analysis. To get a better impression of Q 2 ’s operation, we give examples of how it operates in its various stages. Figure 2 presents an example for an inconsistent response, together with a generated question and the answer Q 2 obtained based on the knowledge. In this example, the ques… find
  4. Conclusion and Future Work We presented Q 2 , an automatic evaluation method for factual consistency in knowledge grounded dialogue. Q 2 employs question generation, question answering and NLI models, and does not require reference responses. To test our approach, we compiled a dataset of dialogue r… find
  5. Evaluation via Question Answering and Question Generation. QA-based evaluation metrics have been proposed as a means for measuring content coverage in text generation tasks. For example, Eyal et al. (2019) used QA models for abstractive summarization both as an evaluation metric and as an optimizati… find
  6. Durmus et al. (2020) and Wang et al. (2020) suggested using QG and QA to identify factual inconsistencies in abstractive summaries, which is more Precision-oriented. Their approach is based on the intuition Acknowledgements that if a summary is consistent with its source, This work was carried out a… find our card
  7. Antoine Bosselut, Esin Durmus, Varun Prashant Gangal, Sebastian Gehrmann, Yacine Jernite, Laura Perez-Beltrachini, Samira Shaikh, and Wei Xu, editors. 2021. Proceedings of the 1st Workshop on Natural Language Generation, Evaluation, and Metrics (GEM 2021). Association for Computational Linguistics,… find
  8. Samuel R. Bowman, Gabor Angeli, Christopher Potts, and Christopher D. Manning. 2015. A large annotated corpus for learning natural language inference. In Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, pages 632–642, Lisbon, Portugal. Association for Computati… find
  9. Angela Fan, Mike Lewis, and Yann Dauphin. 2018. Hierarchical neural story generation. In Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 889–898, Melbourne, Australia. Association for Computational Linguistics. find
  10. Matt Gardner, Joel Grus, Mark Neumann, Oyvind Tafjord, Pradeep Dasigi, Nelson F. Liu, Matthew Peters, Michael Schmitz, and Luke S. Zettlemoyer. 2017. Allennlp: A deep semantic natural language processing platform. find our card
  11. Ben Goodrich, Vinay Rao, Peter J. Liu, and Mohammad Saleh. 2019. Assessing the factual accuracy of generated text. In KDD ’19, KDD ’19, page 166–175, New York, NY, USA. Association for Computing Machinery. find
  12. Asli Celikyilmaz, Elizabeth Clark, and Jianfeng Gao. 2020. Evaluation of text generation: A survey. find
  13. Karthik Gopalakrishnan, Behnam Hedayatnia, Qinlang Chen, Anna Gottardi, Sanjeev Kwatra, Anu Venkatesh, Raefer Gabriel, and Dilek Hakkani-Tür. 2019. find
  14. Topical-Chat: Towards Knowledge-Grounded Open-Domain Conversations. In Proc. Interspeech 2019, pages 1891–1895. find
  15. Ido Dagan, Oren Glickman, and Bernardo Magnini. 2006. The pascal recognising textual entailment challenge. In Machine Learning Challenges. Evaluating Predictive Uncertainty, Visual Object Classification, and Recognising Tectual Entailment, pages 177–190, Berlin, Heidelberg. Springer Berlin Heidelber… find
  16. Yvette Graham and Qun Liu. 2016. Achieving accurate conclusions in evaluation of automatic machine translation metrics. In Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 1–10, San Diego, California… find
  17. Emily Dinan, Stephen Roller, Kurt Shuster, Angela Fan, Michael Auli, and Jason Weston. 2019. Wizard of Wikipedia: Knowledge-powered conversational agents. In Proceedings of the International Conference on Learning Representations (ICLR). find
  18. Ari Holtzman, Jan Buys, Li Du, Maxwell Forbes, and Yejin Choi. 2020. The curious case of neural text degeneration. In International Conference on Learning Representations. find
  19. Ziqiang Cao, Furu Wei, W. Li, and Sujian Li. 2018. Faithful to the original: Fact aware neural abstractive summarization. In AAAI. find
  20. Esin Durmus, He He, and Mona Diab. 2020. FEQA: A question answering evaluation framework for faithfulness assessment in abstractive summarization. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 5055– 5070, Online. Association for Computational Lingu… find our card
  21. Nouha Dziri, Hannah Rashkin, Tal Linzen, and David Reitter. 2021. Evaluating groundedness in dialogue systems: The begin benchmark. find
  22. B. Efron. 1987. Better bootstrap confidence intervals. Journal of the American Statistical Association, 82:171–185. find
  23. Matan Eyal, Tal Baumel, and Michael Elhadad. 2019. Question answering as an automatic evaluation metric for news article summarization. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long an… find
  24. Philipp Koehn and Rebecca Knowles. 2017. Six challenges for neural machine translation. In Proceedings of the First Workshop on Neural Machine Translation, pages 28–39, Vancouver. Association for Computational Linguistics. find
  25. Wojciech Kryściński, Bryan McCann, Caiming Xiong, and Richard Socher. 2019. Evaluating the factual consistency of abstractive text summarization. arXiv preprint arXiv:1910.12840. find
  26. Zhenzhong Lan, Mingda Chen, Sebastian Goodman, Kevin Gimpel, Piyush Sharma, and Radu Soricut. 2020. Albert: A lite bert for self-supervised learning of language representations. In International Conference on Learning Representations. find
  27. Katherine Lee, Orhan Firat, Ashish Agarwal, Clara Fannjiang, and David Sussillo. 2019. Hallucinations in neural machine translation. find
  28. Chin-Yew Lin. 2004. ROUGE: A package for automatic evaluation of summaries. In Text Summarization Branches Out, pages 74–81, Barcelona, Spain. Association for Computational Linguistics. find
  29. Chia-Wei Liu, Ryan Lowe, Iulian Serban, Mike Noseworthy, Laurent Charlin, and Joelle Pineau. 2016. How NOT to evaluate your dialogue system: An empirical study of unsupervised evaluation metrics for dialogue response generation. In Proceedings of the 2016 Conference on Empirical Methods in Natural L… find
  30. Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, and Veselin Stoyanov. 2019. Roberta: A robustly optimized bert pretraining approach. arXiv preprint arXiv:1907.11692. find
  31. Ryan Lowe, Michael Noseworthy, Iulian Vlad Serban, Nicolas Angelard-Gontier, Yoshua Bengio, and Joelle Pineau. 2017. Towards an automatic Turing test: Learning to evaluate dialogue responses. In Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long P… find our card
  32. Joshua Maynez, Shashi Narayan, Bernd Bohnet, and Ryan McDonald. 2020. On faithfulness and factuality in abstractive summarization. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 1906–1919, Online. Association for Computational Linguistics. find
  33. Shikib Mehri and Maxine Eskenazi. 2020. USR: An unsupervised and reference free evaluation metric for dialog generation. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 681–707, Online. Association for Computational Linguistics. find our card
  34. A. H. Miller, W. Feng, A. Fisch, J. Lu, D. Batra, A. Bordes, D. Parikh, and J. Weston. 2017. Parlai: A dialog research software platform. arXiv preprint arXiv:1705.06476. find our card
  35. Mathias Müller, Annette Rios, and Rico Sennrich. 2020. Domain robustness in neural machine translation. In Proceedings of the 14th Conference of the Association for Machine Translation in the Americas (Volume 1: Research Track), pages 151–164, Virtual. Association for Machine Translation in the Amer… find our card
  36. Feng Nan, Cicero Nogueira dos Santos, Henghui Zhu, Patrick Ng, Kathleen McKeown, Ramesh Nallapati, Dejiao Zhang, Zhiguo Wang, Andrew O. Arnold, and Bing Xiang. 2021. Improving factual consistency of abstractive summarization via question answering. In Proceedings of the 59th Annual Meeting of the As… find
  37. Bo Pang, Erik Nijkamp, Wenjuan Han, Linqi Zhou, Yixian Liu, and Kewei Tu. 2020. Towards holistic and automatic evaluation of open-domain dialogue generation. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 3619–3629, Online. Association for Computati… find
  38. Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. 2002. Bleu: a method for automatic evaluation of machine translation. In Proceedings of the 40th Annual Meeting of the Association for Computational Linguistics, pages 311–318, Philadelphia, Pennsylvania, USA. Association for Computational… find our card
  39. Alec Radford, Jeff Wu, Rewon Child, David Luan, Dario Amodei, and Ilya Sutskever. 2019. Language models are unsupervised multitask learners. Technical Report. find
  40. Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J. Liu. 2020. Exploring the limits of transfer learning with a unified text-to-text transformer. Journal of Machine Learning Research, 21(140):1–67. find
  41. Pranav Rajpurkar, Robin Jia, and Percy Liang. 2018. Know what you don’t know: Unanswerable questions for SQuAD. In Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), pages 784– 789, Melbourne, Australia. Association for Computational Lin… find our card
  42. Pranav Rajpurkar, Jian Zhang, Konstantin Lopyrev, and Percy Liang. 2016. SQuAD: 100,000+ questions for machine comprehension of text. In Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, pages 2383–2392, Austin, Texas. Association for Computational Linguistics. find our card
  43. Hannah Rashkin, David Reitter, Gaurav Singh Tomar, and Dipanjan Das. 2021. Increasing faithfulness in knowledge-grounded dialogue with controllable features. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natu… find
  44. Stephen Roller, Emily Dinan, Naman Goyal, Da Ju, Mary Williamson, Yinhan Liu, Jing Xu, Myle Ott, Kurt Shuster, Eric M. Smith, Y-Lan Boureau, and Jason Weston. 2020. Recipes for building an open-domain chatbot. find our card
  45. Thomas Scialom, Paul-Alexis Dray, Gallinari Patrick, Lamprier Sylvain, Piwowarski Benjamin, Staiano Jacopo, and Wang Alex. 2021. Questeval: Summarization asks for fact-based evaluation. arXiv preprint arXiv:2103.12693. find our card
  46. Thomas Scialom, Sylvain Lamprier, Benjamin Piwowarski, and Jacopo Staiano. 2019. Answers unite! unsupervised metrics for reinforced summarization models. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural La… find our card
  47. Thibault Sellam, Dipanjan Das, and Ankur Parikh. 2020. BLEURT: Learning robust metrics for text generation. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 7881–7892, Online. Association for Computational Linguistics. find
  48. Linguistics (Volume 1: Long Papers), pages 2204– 2213, Melbourne, Australia. Association for Computational Linguistics. find
  49. Tianyi Zhang, Varsha Kishore, Felix Wu, Kilian Q. Weinberger, and Yoav Artzi. 2020. Bertscore: Evaluating text generation with bert. In International Conference on Learning Representations. find
  50. Tianyu Zhao, Divesh Lala, and Tatsuya Kawahara. 2020. Designing precise and robust dialogue response evaluators. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 26–33, Online. Association for Computational Linguistics. find
  51. Kurt Shuster, Da Ju, Stephen Roller, Emily Dinan, Y-Lan Boureau, and Jason Weston. 2020. The dialogue dodecathlon: Open-domain knowledge and image grounded conversational agents. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 2453–2470, Online. Asso… find
  52. James Thorne and Andreas Vlachos. 2018. Automated fact checking: Task formulations, methods and future directions. In Proceedings of the 27th International Conference on Computational Linguistics, pages 3346–3359, Santa Fe, New Mexico, USA. Association for Computational Linguistics. find our card
  53. Alex Wang, Kyunghyun Cho, and Mike Lewis. 2020. Asking and answering questions to evaluate the factual consistency of summaries. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 5008–5020, Online. Association for Computational Linguistics. find
  54. Chaojun Wang and Rico Sennrich. 2020. On exposure bias, hallucination and domain shift in neural machine translation. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 3544–3552, Online. Association for Computational Linguistics. find our card
  55. Sean Welleck, Jason Weston, Arthur Szlam, and Kyunghyun Cho. 2019. Dialogue natural language inference. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 3731–3741, Florence, Italy. Association for Computational Linguistics. find our card
  56. Xinnuo Xu, Ondřej Dušek, Jingyi Li, Verena Rieser, and Ioannis Konstas. 2020. Fact-based content weighting for evaluating abstractive summarisation. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 5071–5081, Online. Association for Computational Lin… find
  57. Saizheng Zhang, Emily Dinan, Jack Urbanek, Arthur Szlam, Douwe Kiela, and Jason Weston. 2018. Personalizing dialogue agents: I have a dog, do you have pets too? In Proceedings of the 56th Annual Meeting of the Association for Computational find our card

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/honovich-2021-q-2-evaluating-factual-consistency/