Multi-Granularity Optimization for Non-Autoregressive Translation

TL;DR

This paper introduces multi-granularity optimization for non-autoregressive translation, improving performance by incorporating feedback from translation segments of various granularities, thus addressing the limitations of the independence assumption.

Contribution

It proposes a novel multi-granularity optimization method that enhances non-autoregressive translation by leveraging feedback from multiple translation segment levels.

Findings

Significant performance improvements over baseline models

Best results on WMT'16 En-Ro translation task

Highly competitive results on WMT'14 En-De

Abstract

Despite low latency, non-autoregressive machine translation (NAT) suffers severe performance deterioration due to the naive independence assumption. This assumption is further strengthened by cross-entropy loss, which encourages a strict match between the hypothesis and the reference token by token. To alleviate this issue, we propose multi-granularity optimization for NAT, which collects model behaviors on translation segments of various granularities and integrates feedback for backpropagation. Experiments on four WMT benchmarks show that the proposed method significantly outperforms the baseline models trained with cross-entropy loss, and achieves the best performance on WMT'16 En-Ro and highly competitive results on WMT'14 En-De for fully non-autoregressive translation.