Incremental Blockwise Beam Search for Simultaneous Speech Translation with Controllable Quality-Latency Tradeoff

TL;DR

This paper introduces an incremental blockwise beam search method for simultaneous speech translation that allows for controllable quality-latency tradeoffs, improving translation quality or reducing latency without sacrificing the other.

Contribution

It proposes a novel incremental beam search approach with quality-latency control mechanisms applicable to online and offline models, enhancing real-time translation performance.

Findings

Achieved 0.6-3.6 BLEU improvement without latency increase

Reduced latency by 0.8-1.4 seconds without quality loss

Effective application to both online and offline translation models

Abstract

Blockwise self-attentional encoder models have recently emerged as one promising end-to-end approach to simultaneous speech translation. These models employ a blockwise beam search with hypothesis reliability scoring to determine when to wait for more input speech before translating further. However, this method maintains multiple hypotheses until the entire speech input is consumed -- this scheme cannot directly show a single \textit{incremental} translation to users. Further, this method lacks mechanisms for \textit{controlling} the quality vs. latency tradeoff. We propose a modified incremental blockwise beam search incorporating local agreement or hold-$n$ policies for quality-latency control. We apply our framework to models trained for online or offline translation and demonstrate that both types can be effectively used in online mode. Experimental results on MuST-C show 0.6-3.6 BLEU improvement without changing latency or 0.8-1.4 s latency improvement without changing quality.