Adaptive Policy with Wait-$k$ Model for Simultaneous Translation
This work addresses the challenge of efficient and accurate real-time translation for applications like live captioning, though it is incremental as it builds on existing wait-k and adaptive policy methods.
The paper tackles the problem of simultaneous machine translation by proposing a divergence-based adaptive policy (DaP) that decouples the policy from the translation model, using a frozen wait-k model with lightweight parameters. Experimental results show it offers an improved trade-off between translation accuracy and latency, outperforming strong baselines across various benchmarks.
Simultaneous machine translation (SiMT) requires a robust read/write policy in conjunction with a high-quality translation model. Traditional methods rely on either a fixed wait-$k$ policy coupled with a standalone wait-$k$ translation model, or an adaptive policy jointly trained with the translation model. In this study, we propose a more flexible approach by decoupling the adaptive policy model from the translation model. Our motivation stems from the observation that a standalone multi-path wait-$k$ model performs competitively with adaptive policies utilized in state-of-the-art SiMT approaches. Specifically, we introduce DaP, a divergence-based adaptive policy, that makes read/write decisions for any translation model based on the potential divergence in translation distributions resulting from future information. DaP extends a frozen wait-$k$ model with lightweight parameters, and is both memory and computation efficient. Experimental results across various benchmarks demonstrate that our approach offers an improved trade-off between translation accuracy and latency, outperforming strong baselines.