Data-Driven Adaptive Simultaneous Machine Translation

TL;DR

This paper introduces an adaptive training scheme for simultaneous machine translation that improves over traditional fixed policies by allowing dynamic latency adjustment without increasing training complexity.

Contribution

It proposes a novel training method that augments data with adaptive prefix pairs, enabling more flexible and efficient SimulMT training.

Findings

Outperforms strong baselines in translation quality

Achieves better latency-quality trade-offs

Maintains training complexity similar to full-sentence models

Abstract

In simultaneous translation (SimulMT), the most widely used strategy is the wait-k policy thanks to its simplicity and effectiveness in balancing translation quality and latency. However, wait-k suffers from two major limitations: (a) it is a fixed policy that can not adaptively adjust latency given context, and (b) its training is much slower than full-sentence translation. To alleviate these issues, we propose a novel and efficient training scheme for adaptive SimulMT by augmenting the training corpus with adaptive prefix-to-prefix pairs, while the training complexity remains the same as that of training full-sentence translation models. Experiments on two language pairs show that our method outperforms all strong baselines in terms of translation quality and latency.