With a Little Help from the Authors: Reproducing Human Evaluation of an MT Error Detector

TL;DR

This paper attempts to reproduce a human evaluation study of an MT error detector, confirming its main conclusions but highlighting variability in human annotations and reproducibility challenges.

Contribution

It provides a detailed reproduction of a previous human evaluation experiment and discusses reproducibility issues and variability in human annotations.

Findings

Replicated results generally confirm original conclusions

Identified high variability in human annotation

Highlighted reproducibility challenges in human evaluation

Abstract

This work presents our efforts to reproduce the results of the human evaluation experiment presented in the paper of Vamvas and Sennrich (2022), which evaluated an automatic system detecting over- and undertranslations (translations containing more or less information than the original) in machine translation (MT) outputs. Despite the high quality of the documentation and code provided by the authors, we discuss some problems we found in reproducing the exact experimental setup and offer recommendations for improving reproducibility. Our replicated results generally confirm the conclusions of the original study, but in some cases, statistically significant differences were observed, suggesting a high variability of human annotation.