Alignatt: Using Attention-based Audio-translation Alignments As A Guide For Simultaneous Speech Translation

Abstract

Attention is the core mechanism of today's most used architectures for natural language processing and has been analyzed from many perspectives, including its effectiveness for machine translation-related tasks. Among these studies, attention resulted to be a useful source of information to get insights about word alignment also when the input text is substituted with audio segments, as in the case of the speech translation (ST) task. In this paper, we propose AlignAtt, a novel policy for simultaneous ST (SimulST) that exploits the attention information to generate source-target alignments that guide the model during inference. Through experiments on the 8 language pairs of MuST-C v1.0, we show that AlignAtt outperforms previous state-of-the-art SimulST policies applied to offline-trained models with gains in terms of BLEU of 2 points and latency reductions ranging from 0.5s to 0.8s across the 8 languages.

Stats

• citations2
• S2 citations—
• github stars0
• HF likes0
• heat score3.58
• arxiv keypapi2023alignatt

Related papers

• Streamatt: Direct Streaming Speech-to-text Translation With Attention-based Audio History Selection (2024)4.52
• Sample, Translate, Recombine: Leveraging Audio Alignments For Data Augmentation In End-to-end Speech Translation (2022)7.81
• Efficient And Adaptive Simultaneous Speech Translation With Fully Unidirectional Architecture (2025)2.26
• Speechalign: A Framework For Speech Translation Alignment Evaluation (2023)0.00
• Decision Attentive Regularization To Improve Simultaneous Speech Translation Systems (2021)0.00
• Synchronous Speech Recognition And Speech-to-text Translation With Interactive Decoding (2019)10.48
• Exploring Continuous Integrate-and-fire For Adaptive Simultaneous Speech Translation (2022)4.52
• Learning When To Speak: Latency And Quality Trade-offs For Simultaneous Speech-to-speech Translation With Offline Models (2023)0.00