Monotonic Chunkwise Attention
This addresses the inefficiency of soft attention for real-time sequence transduction, though it is incremental as it builds on monotonic attention methods.
The authors tackled the quadratic time and space cost of soft attention in sequence-to-sequence models by proposing Monotonic Chunkwise Attention (MoChA), which enables online and linear-time decoding while achieving state-of-the-art results in online speech recognition and improved performance in document summarization.
Sequence-to-sequence models with soft attention have been successfully applied to a wide variety of problems, but their decoding process incurs a quadratic time and space cost and is inapplicable to real-time sequence transduction. To address these issues, we propose Monotonic Chunkwise Attention (MoChA), which adaptively splits the input sequence into small chunks over which soft attention is computed. We show that models utilizing MoChA can be trained efficiently with standard backpropagation while allowing online and linear-time decoding at test time. When applied to online speech recognition, we obtain state-of-the-art results and match the performance of a model using an offline soft attention mechanism. In document summarization experiments where we do not expect monotonic alignments, we show significantly improved performance compared to a baseline monotonic attention-based model.