Missing information, unresponsive authors, experimental flaws : the impossibility of assessing the reproducibility of previous human evaluations in NLP

Belz, Anya; Thomson, Craig; Reiter, Ehud; Cieliebak, Mark; Hürlimann, Manuela

doi:10.18653/v1/2023.insights-1.1

Please use this identifier to cite or link to this item: https://doi.org/10.21256/zhaw-29064

Publication type:	Conference paper
Type of review:	Peer review (publication)
Title:	Missing information, unresponsive authors, experimental flaws : the impossibility of assessing the reproducibility of previous human evaluations in NLP
Authors:	Belz, Anya Thomson, Craig Reiter, Ehud Cieliebak, Mark Hürlimann, Manuela
et. al:	Yes
DOI:	10.18653/v1/2023.insights-1.1 10.21256/zhaw-29064
Proceedings:	The Fourth Workshop on Insights from Negative Results in NLP
Page(s):	1
Pages to:	10
Conference details:	The Fourth Workshop on Insights from Negative Results in NLP, Dubrovnik, Croatia, 2-6 May 2023
Issue Date:	2023
Publisher / Ed. Institution:	Association for Computational Linguistics
Language:	English
Subjects:	Reproducibility; System evaluation; Natural language processing
Subject (DDC):	410.285: Computational linguistics
Abstract:	We report our efforts in identifying a set of previous human evaluations in NLP that would be suitable for a coordinated study examining what makes human evaluations in NLP more/less reproducible. We present our results and findings, which include that just 13% of papers had (i) sufficiently low barriers to reproduction, and (ii) enough obtainable information, to be considered for reproduction, and that all but one of the experiments we selected for reproduction was discovered to have flaws that made the meaningfulness of conducting a reproduction questionable. As a result, we had to change our coordinated study design from a reproduce approach to a standardise-then-reproduce-twice approach. Our overall (negative) finding that the great majority of human evaluations in NLP is not repeatable and/or not reproducible and/or too flawed to justify reproduction, paints a dire picture, but presents an opportunity for a rethink about how to design and report human evaluations in NLP.
URI:	https://digitalcollection.zhaw.ch/handle/11475/29064
Fulltext version:	Published version
License (according to publishing contract):	CC BY 4.0: Attribution 4.0 International
Departement:	School of Engineering
Organisational Unit:	Centre for Artificial Intelligence (CAI)
Appears in collections:	Publikationen School of Engineering

Files in This Item:

File	Description	Size	Format
2023_Belz-etal_Reproducibility-of-previous-human-evaluations-in-NLP.pdf		145.07 kB	Adobe PDF	View/Open

Show full item record

Belz, A., Thomson, C., Reiter, E., Cieliebak, M., & Hürlimann, M. (2023). Missing information, unresponsive authors, experimental flaws : the impossibility of assessing the reproducibility of previous human evaluations in NLP [Conference paper]. The Fourth Workshop on Insights from Negative Results in NLP, 1–10. https://doi.org/10.18653/v1/2023.insights-1.1

Belz, A. et al. (2023) ‘Missing information, unresponsive authors, experimental flaws : the impossibility of assessing the reproducibility of previous human evaluations in NLP’, in The Fourth Workshop on Insights from Negative Results in NLP. Association for Computational Linguistics, pp. 1–10. Available at: https://doi.org/10.18653/v1/2023.insights-1.1.

A. Belz, C. Thomson, E. Reiter, M. Cieliebak, and M. Hürlimann, “Missing information, unresponsive authors, experimental flaws : the impossibility of assessing the reproducibility of previous human evaluations in NLP,” in The Fourth Workshop on Insights from Negative Results in NLP, 2023, pp. 1–10. doi: 10.18653/v1/2023.insights-1.1.

BELZ, Anya, Craig THOMSON, Ehud REITER, Mark CIELIEBAK und Manuela HÜRLIMANN, 2023. Missing information, unresponsive authors, experimental flaws : the impossibility of assessing the reproducibility of previous human evaluations in NLP. In: The Fourth Workshop on Insights from Negative Results in NLP. Conference paper. Association for Computational Linguistics. 2023. S. 1–10

Belz, Anya, Craig Thomson, Ehud Reiter, Mark Cieliebak, and Manuela Hürlimann. 2023. “Missing Information, Unresponsive Authors, Experimental Flaws : The Impossibility of Assessing the Reproducibility of Previous Human Evaluations in NLP.” Conference paper. In The Fourth Workshop on Insights from Negative Results in NLP, 1–10. Association for Computational Linguistics. https://doi.org/10.18653/v1/2023.insights-1.1.

Belz, Anya, et al. “Missing Information, Unresponsive Authors, Experimental Flaws : The Impossibility of Assessing the Reproducibility of Previous Human Evaluations in NLP.” The Fourth Workshop on Insights from Negative Results in NLP, Association for Computational Linguistics, 2023, pp. 1–10, https://doi.org/10.18653/v1/2023.insights-1.1.