AmericasNLI: Evaluating Zero-shot Natural Language Understanding of Pretrained Multilingual Models in Truly Low-resource Languages

TL;DR

This paper evaluates the zero-shot semantic understanding capabilities of pretrained multilingual models on 10 indigenous American languages using the AmericasNLI dataset, revealing limited initial performance but improvements through continued pretraining and translation-based methods.

Contribution

It introduces AmericasNLI for low-resource languages and compares zero-shot, translation, and adaptation approaches, highlighting the challenges and potential strategies for semantic tasks in unseen languages.

Findings

XLM-R's zero-shot accuracy is around 38.62% for the languages.

Continued pretraining improves accuracy to 44.05%.

Training on poorly translated data surpasses other methods with 48.72% accuracy.

Abstract

Pretrained multilingual models are able to perform cross-lingual transfer in a zero-shot setting, even for languages unseen during pretraining. However, prior work evaluating performance on unseen languages has largely been limited to low-level, syntactic tasks, and it remains unclear if zero-shot learning of high-level, semantic tasks is possible for unseen languages. To explore this question, we present AmericasNLI, an extension of XNLI (Conneau et al., 2018) to 10 indigenous languages of the Americas. We conduct experiments with XLM-R, testing multiple zero-shot and translation-based approaches. Additionally, we explore model adaptation via continued pretraining and provide an analysis of the dataset by considering hypothesis-only models. We find that XLM-R's zero-shot performance is poor for all 10 languages, with an average performance of 38.62%. Continued pretraining offers improvements, with an average accuracy of 44.05%. Surprisingly, training on poorly translated data by far outperforms all other methods with an accuracy of 48.72%.