Multilingual Transformer Encoders: a Word-Level Task-Agnostic Evaluation
This work addresses the need for better evaluation of cross-lingual transfer in NLP, but it is incremental as it builds on existing methods for alignment assessment.
The paper tackled the problem of evaluating whether multilingual Transformer models learn universal cross-lingual patterns by proposing a word-level task-agnostic method, which provided more accurate translated word pairs and showed that some inner layers outperform other aligned representations, especially under stricter alignment definitions.
Some Transformer-based models can perform cross-lingual transfer learning: those models can be trained on a specific task in one language and give relatively good results on the same task in another language, despite having been pre-trained on monolingual tasks only. But, there is no consensus yet on whether those transformer-based models learn universal patterns across languages. We propose a word-level task-agnostic method to evaluate the alignment of contextualized representations built by such models. We show that our method provides more accurate translated word pairs than previous methods to evaluate word-level alignment. And our results show that some inner layers of multilingual Transformer-based models outperform other explicitly aligned representations, and even more so according to a stricter definition of multilingual alignment.