CoSDA-ML: Multi-Lingual Code-Switching Data Augmentation for Zero-Shot Cross-Lingual NLP
This work addresses a key limitation in zero-shot cross-lingual NLP for researchers and practitioners, offering a more efficient method that does not rely on bilingual sentences and supports multiple target languages in one training process.
The paper tackles the problem of inconsistent contextualized representations across languages in multilingual models like mBERT by proposing a data augmentation framework that generates multilingual code-switching data for fine-tuning, resulting in significantly improved performances on five tasks across 19 languages compared to mBERT.
Multi-lingual contextualized embeddings, such as multilingual-BERT (mBERT), have shown success in a variety of zero-shot cross-lingual tasks. However, these models are limited by having inconsistent contextualized representations of subwords across different languages. Existing work addresses this issue by bilingual projection and fine-tuning technique. We propose a data augmentation framework to generate multi-lingual code-switching data to fine-tune mBERT, which encourages model to align representations from source and multiple target languages once by mixing their context information. Compared with the existing work, our method does not rely on bilingual sentences for training, and requires only one training process for multiple target languages. Experimental results on five tasks with 19 languages show that our method leads to significantly improved performances for all the tasks compared with mBERT.