Shiftable Context: Addressing Training-Inference Context Mismatch in Simultaneous Speech Translation
This addresses a specific bottleneck for simultaneous speech translation systems, offering an incremental improvement to enhance translation accuracy in streaming tasks.
The paper tackled the training-inference context mismatch problem in simultaneous speech translation by proposing Shiftable Context, a scheme that maintains consistent segment and context sizes, resulting in average BLEU score increases of 1.83 to 2.09 across three language pairs with minimal impact on lagging.
Transformer models using segment-based processing have been an effective architecture for simultaneous speech translation. However, such models create a context mismatch between training and inference environments, hindering potential translation accuracy. We solve this issue by proposing Shiftable Context, a simple yet effective scheme to ensure that consistent segment and context sizes are maintained throughout training and inference, even with the presence of partially filled segments due to the streaming nature of simultaneous translation. Shiftable Context is also broadly applicable to segment-based transformers for streaming tasks. Our experiments on the English-German, English-French, and English-Spanish language pairs from the MUST-C dataset demonstrate that when applied to the Augmented Memory Transformer, a state-of-the-art model for simultaneous speech translation, the proposed scheme achieves an average increase of 2.09, 1.83, and 1.95 BLEU scores across each wait-k value for the three language pairs, respectively, with a minimal impact on computation-aware Average Lagging.