Adapting Offline Speech Translation Models for Streaming with Future-Aware Distillation and Inference
This work addresses a practical problem for real-time speech translation systems by adapting offline models for streaming use, though it is incremental as it builds on existing wait-k policies.
The paper tackled the mismatch between offline-trained speech translation models and streaming inference by proposing Future-Aware Streaming Translation (FAST), which improved translation quality-latency trade-offs on MuST-C benchmarks, achieving better performance than strong baselines.
A popular approach to streaming speech translation is to employ a single offline model with a wait-k policy to support different latency requirements, which is simpler than training multiple online models with different latency constraints. However, there is a mismatch problem in using a model trained with complete utterances for streaming inference with partial input. We demonstrate that speech representations extracted at the end of a streaming input are significantly different from those extracted from a complete utterance. To address this issue, we propose a new approach called Future-Aware Streaming Translation (FAST) that adapts an offline ST model for streaming input. FAST includes a Future-Aware Inference (FAI) strategy that incorporates future context through a trainable masked embedding, and a Future-Aware Distillation (FAD) framework that transfers future context from an approximation of full speech to streaming input. Our experiments on the MuST-C EnDe, EnEs, and EnFr benchmarks show that FAST achieves better trade-offs between translation quality and latency than strong baselines. Extensive analyses suggest that our methods effectively alleviate the aforementioned mismatch problem between offline training and online inference.