Data-Efficient Cross-Lingual Transfer with Language-Specific Subnetworks
This addresses data efficiency in multilingual NLP, but it is incremental as it builds on existing meta-learning techniques.
The paper tackled the problem of conflicting training updates across languages in multilingual models by proposing language-specific subnetworks to control parameter sharing, resulting in improved cross-lingual transfer during fine-tuning.
Large multilingual language models typically share their parameters across all languages, which enables cross-lingual task transfer, but learning can also be hindered when training updates from different languages are in conflict. In this paper, we propose novel methods for using language-specific subnetworks, which control cross-lingual parameter sharing, to reduce conflicts and increase positive transfer during fine-tuning. We introduce dynamic subnetworks, which are jointly updated with the model, and we combine our methods with meta-learning, an established, but complementary, technique for improving cross-lingual transfer. Finally, we provide extensive analyses of how each of our methods affects the models.