Language Graph Distillation for Low-Resource Machine Translation
This addresses the problem of limited bilingual data for low-resource machine translation, offering an incremental improvement over existing multilingual knowledge transfer methods.
The paper tackles the challenge of neural machine translation for low-resource languages by proposing a Language Graph and a graph distillation algorithm, which improves translation accuracy by over 3.13 BLEU points on the TED talks dataset.
Neural machine translation on low-resource language is challenging due to the lack of bilingual sentence pairs. Previous works usually solve the low-resource translation problem with knowledge transfer in a multilingual setting. In this paper, we propose the concept of Language Graph and further design a novel graph distillation algorithm that boosts the accuracy of low-resource translations in the graph with forward and backward knowledge distillation. Preliminary experiments on the TED talks multilingual dataset demonstrate the effectiveness of our proposed method. Specifically, we improve the low-resource translation pair by more than 3.13 points in terms of BLEU score.