The Effect of Domain and Diacritics in Yorùbá-English Neural Machine Translation
This work addresses the problem of evaluating machine translation for low-resource languages like Yorùbá, providing a benchmark for future research, though it is incremental as it builds on existing multilingual models.
The paper tackles the lack of standardized evaluation datasets for low-resource language pairs in machine translation by introducing MENYO-20k, a multi-domain parallel corpus for Yorùbá-English, and shows that their models outperform massively multilingual models with improvements of +8.7 and +9.1 BLEU scores when translating to Yorùbá.
Massively multilingual machine translation (MT) has shown impressive capabilities, including zero and few-shot translation between low-resource language pairs. However, these models are often evaluated on high-resource languages with the assumption that they generalize to low-resource ones. The difficulty of evaluating MT models on low-resource pairs is often due to lack of standardized evaluation datasets. In this paper, we present MENYO-20k, the first multi-domain parallel corpus with a special focus on clean orthography for Yorùbá--English with standardized train-test splits for benchmarking. We provide several neural MT benchmarks and compare them to the performance of popular pre-trained (massively multilingual) MT models both for the heterogeneous test set and its subdomains. Since these pre-trained models use huge amounts of data with uncertain quality, we also analyze the effect of diacritics, a major characteristic of Yorùbá, in the training data. We investigate how and when this training condition affects the final quality and intelligibility of a translation. Our models outperform massively multilingual models such as Google ($+8.7$ BLEU) and Facebook M2M ($+9.1$ BLEU) when translating to Yorùbá, setting a high quality benchmark for future research.