How to develop a single unified model to translate from any language to any language? This work proposes a many-to-many translation system with emphasis on both English-centric and non-English directions. Many recent works have focused on proposing a single unified model for multiligual translation. These models are favorable because they are efficient and easy for deployment. However, most of these works focus on improving English-centric directions, which means that translation between two arbitrary languages may not be well supported. Therefore, in this paper, they propose a training method called mRASP2, including contrastive learning and alignment augmentation (AA) to train a unified multilingual translation system. They also contribute a monolingual dataset called MC24. By making use of monolingual and bilingual language copora, the system is able to learn language-agnostic representation to support non-English directions better than before. Their system achieves great performances and outperforms a strong Transformer baseline by a large margin.
- Multilingual MT5
- Pre-training5
- BERT4
- Random Aligned Substitution2
- Zero-shot Translation2
- mRASP2
- Speech Translation2
- MT Evaluation2
- Transformer2
- Knowledge Assessment1
- Risk Ratio1
- Vocabulary Learning1
- Optimal Transport1
- Catastrophic Forgetting1
- Model Protection1
- Relation Extraction1
- Fact Verafication1
- Reasoning1
- Logic-regularized neural network1
- Embedding1
- Shared Semantic Memory1
- Chimera1
- Language Modelling1
- Imagination1
- Visual Machine Translation1
- ImagiT1
- Model Capacity1
- Language-specific Sub-network1
- GPU Acceleration1
- CUDA1
- Variational Inference1
- Latent Variable Model1
- Semi-supervised Learning1
- Contrastive Learning1
- BERTScore1
- COMET1
- Translation Memory1
- Recurrent Attention1
- Self-training1
- Unsupervised Machine Translation1
About 5 min