PsFuture: A Pseudo-Future-based Zero-Shot Adaptive Policy for Simultaneous Machine Translation

TL;DR

PsFuture introduces a zero-shot adaptive policy for simultaneous translation that eliminates the need for additional training, leveraging a novel training strategy to improve quality-latency trade-offs.

Contribution

The paper presents PsFuture, the first zero-shot adaptive read/write policy for SiMT, and a new Prefix-to-Full training method for offline models adapted to real-time translation.

Findings

Zero-shot policy performs on par with strong baselines.

P2F training enhances translation quality and latency trade-off.

Method reduces training complexity for SiMT systems.

Abstract

Simultaneous Machine Translation (SiMT) requires target tokens to be generated in real-time as streaming source tokens are consumed. Traditional approaches to SiMT typically require sophisticated architectures and extensive parameter configurations for training adaptive read/write policies, which in turn demand considerable computational power and memory. We propose PsFuture, the first zero-shot adaptive read/write policy for SiMT, enabling the translation model to independently determine read/write actions without the necessity for additional training. Furthermore, we introduce a novel training strategy, Prefix-to-Full (P2F), specifically tailored to adjust offline translation models for SiMT applications, exploiting the advantages of the bidirectional attention mechanism inherent in offline models. Experiments across multiple benchmarks demonstrate that our zero-shot policy attains performance on par with strong baselines and the P2F method can further enhance performance, achieving an outstanding trade-off between translation quality and latency.