Existing zero-shot cross-lingual transfer methods rely on parallel corpora or bilingual dictionaries, which are expensive and impractical for low-resource languages. To disengage from these dependencies, researchers have explored training multilingual models on English-only resources and transferring them to low-resource languages. However, its effect is limited by the gap between embedding clusters of different languages. To address this issue, we propose Embedding-Push, Attention-Pull, and Robust targets to transfer English embeddings to virtual multilingual embeddings without semantic loss, thereby improving cross-lingual transferability. Experimental results on mBERT and XLM-R demonstrate that our method significantly outperforms previous works on the zero-shot cross-lingual text classification task and can obtain a better multilingual alignment.
@article{arxiv.2210.09934,
title = {A Simple and Effective Method to Improve Zero-Shot Cross-Lingual Transfer Learning},
author = {Kunbo Ding and Weijie Liu and Yuejian Fang and Weiquan Mao and Zhe Zhao and Tao Zhu and Haoyan Liu and Rong Tian and Yiren Chen},
journal= {arXiv preprint arXiv:2210.09934},
year = {2022}
}