What happens if you treat ordinal ratings as interval data? Human evaluations in {NLP} are even more under-powered than you think

Howcroft, David M.; Rieser, Verena

Abstract

Previous work has shown that human evaluations in NLP are notoriously under-powered. Here, we argue that there are two common factors which make this problem even worse: NLP studies usually (a) treat ordinal data as interval data and (b) operate under high variance settings while the differences they are hoping to detect are often subtle. We demonstrate through simulation that ordinal mixed effects models are better able to detect small differences between models, especially in high variance settings common in evaluations of generated texts. We release tools for researchers to conduct their own power analysis and test their assumptions. We also make recommendations for improving statistical power.

Citation

Howcroft, D. M., & Rieser, V. (2021). What happens if you treat ordinal ratings as interval data? Human evaluations in {NLP} are even more under-powered than you think. In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing (8932-8939)

• Conference Name: 2021 Conference on Empirical Methods in Natural Language Processing
• Start Date: Nov 7, 2021
• End Date: Nov 11, 2021
• Acceptance Date: Aug 26, 2021
• Publication Date: 2021-11
• Deposit Date: Dec 2, 2021
• Publicly Available Date: Dec 2, 2021
• Publisher: Association for Computational Linguistics (ACL)
• Pages: 8932-8939
• Book Title: Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing
• Public URL: http://researchrepository.napier.ac.uk/Output/2826175
• Publisher URL: https://aclanthology.org/2021.emnlp-main.703