Evaluating the Cross-Lingual Effectiveness of Massively Multilingual Neural Machine Translation
This work addresses the problem of cross-lingual transfer for downstream tasks, particularly for low-resource languages, but is incremental as it builds on existing multilingual models.
The paper evaluated the cross-lingual effectiveness of representations from a massively multilingual neural machine translation model on 5 downstream tasks covering over 50 languages, showing gains in zero-shot transfer in 4 out of 5 tasks compared to multilingual BERT.
The recently proposed massively multilingual neural machine translation (NMT) system has been shown to be capable of translating over 100 languages to and from English within a single model. Its improved translation performance on low resource languages hints at potential cross-lingual transfer capability for downstream tasks. In this paper, we evaluate the cross-lingual effectiveness of representations from the encoder of a massively multilingual NMT model on 5 downstream classification and sequence labeling tasks covering a diverse set of over 50 languages. We compare against a strong baseline, multilingual BERT (mBERT), in different cross-lingual transfer learning scenarios and show gains in zero-shot transfer in 4 out of these 5 tasks.