Translation Errors Significantly Impact Low-Resource Languages in Cross-Lingual Learning
This work highlights a critical data quality issue that skews evaluation for low-resource languages in cross-lingual learning, which is incremental as it builds on existing benchmark analysis.
The paper identifies that translation inconsistencies in cross-lingual benchmarks like XNLI disproportionately affect low-resource languages, and it proposes a method to detect these errors by comparing performance gaps between human-translated and machine-translated text, with manual reannotation confirming errors in Hindi and Urdu.
Popular benchmarks (e.g., XNLI) used to evaluate cross-lingual language understanding consist of parallel versions of English evaluation sets in multiple target languages created with the help of professional translators. When creating such parallel data, it is critical to ensure high-quality translations for all target languages for an accurate characterization of cross-lingual transfer. In this work, we find that translation inconsistencies do exist and interestingly they disproportionally impact low-resource languages in XNLI. To identify such inconsistencies, we propose measuring the gap in performance between zero-shot evaluations on the human-translated and machine-translated target text across multiple target languages; relatively large gaps are indicative of translation errors. We also corroborate that translation errors exist for two target languages, namely Hindi and Urdu, by doing a manual reannotation of human-translated test instances in these two languages and finding poor agreement with the original English labels these instances were supposed to inherit.