The USTC-NELSLIP Systems for Simultaneous Speech Translation Task at IWSLT 2021

TL;DR

This paper introduces the CAAT model for simultaneous speech translation, demonstrating improved quality-latency trade-offs over previous methods through experiments on speech-to-text and text-to-text tasks.

Contribution

We propose the Cross Attention Augmented Transducer (CAAT), a novel model extending RNN-T for sequence-to-sequence simultaneous translation without monotonic constraints.

Findings

CAAT outperforms wait-k in quality-latency trade-offs.

S2T system improves BLEU by 11.3 points over previous systems.

T2T system improves BLEU by 4.6 points over previous systems.

Abstract

This paper describes USTC-NELSLIP's submissions to the IWSLT2021 Simultaneous Speech Translation task. We proposed a novel simultaneous translation model, Cross Attention Augmented Transducer (CAAT), which extends conventional RNN-T to sequence-to-sequence tasks without monotonic constraints, e.g., simultaneous translation. Experiments on speech-to-text (S2T) and text-to-text (T2T) simultaneous translation tasks shows CAAT achieves better quality-latency trade-offs compared to \textit{wait-k}, one of the previous state-of-the-art approaches. Based on CAAT architecture and data augmentation, we build S2T and T2T simultaneous translation systems in this evaluation campaign. Compared to last year's optimal systems, our S2T simultaneous translation system improves by an average of 11.3 BLEU for all latency regimes, and our T2T simultaneous translation system improves by an average of 4.6 BLEU.