StreamAtt: Direct Streaming Speech-to-Text Translation with Attention-based Audio History Selection

About

Streaming speech-to-text translation (StreamST) is the task of automatically translating speech while incrementally receiving an audio stream. Unlike simultaneous ST (SimulST), which deals with pre-segmented speech, StreamST faces the challenges of handling continuous and unbounded audio streams. This requires additional decisions about what to retain of the previous history, which is impractical to keep entirely due to latency and computational constraints. Despite the real-world demand for real-time ST, research on streaming translation remains limited, with existing works solely focusing on SimulST. To fill this gap, we introduce StreamAtt, the first StreamST policy, and propose StreamLAAL, the first StreamST latency metric designed to be comparable with existing metrics for SimulST. Extensive experiments across all 8 languages of MuST-C v1.0 show the effectiveness of StreamAtt compared to a naive streaming baseline and the related state-of-the-art SimulST policy, providing a first step in StreamST research.

Sara Papi, Marco Gaido, Matteo Negri, Luisa Bentivogli · 2024

Related benchmarks

Task	Dataset	Result
Latency Metric Accuracy Evaluation	Long-form SimulST (En-De)	LAAL (Stream)92	1
Latency Metric Accuracy Evaluation	Long-form SimulST En-Zh	Stream LAAL86	1
Latency Metric Accuracy Evaluation	Long-form SimulST En-Ja	Stream LAAL0.63	1
Latency Metric Accuracy Evaluation	Long-form SimulST Cs-En	LAAL (Stream)77	1
Latency Metric Accuracy Evaluation	Long-form SimulST Different Team	Stream LAAL82	1
Latency Metric Accuracy Evaluation	Long-form SimulST Same Team	Stream LAAL0.83	1
Latency Metric Accuracy Evaluation	Long-form SimulST All language pairs	AL—	1