Revisiting Modularized Multilingual NMT to Meet Industrial Demands
This addresses industrial needs for more maintainable and scalable multilingual translation systems, though it is incremental as it builds on existing modular methods.
The study tackled the performance degradation and low maintainability of fully shared multilingual NMT models by revisiting a modular approach that shares modules only among same languages, finding it avoids capacity bottlenecks and allows incremental module addition with zero-shot performance comparable to supervised models.
The complete sharing of parameters for multilingual translation (1-1) has been the mainstream approach in current research. However, degraded performance due to the capacity bottleneck and low maintainability hinders its extensive adoption in industries. In this study, we revisit the multilingual neural machine translation model that only share modules among the same languages (M2) as a practical alternative to 1-1 to satisfy industrial requirements. Through comprehensive experiments, we identify the benefits of multi-way training and demonstrate that the M2 can enjoy these benefits without suffering from the capacity bottleneck. Furthermore, the interlingual space of the M2 allows convenient modification of the model. By leveraging trained modules, we find that incrementally added modules exhibit better performance than singly trained models. The zero-shot performance of the added modules is even comparable to supervised models. Our findings suggest that the M2 can be a competent candidate for multilingual translation in industries.