Deep Sparse Conformer for Speech Recognition
This work addresses speech recognition efficiency and accuracy for Japanese language processing, representing an incremental improvement over existing Conformer models.
The paper tackles improving Conformer's long-sequence representation for speech recognition by making it sparser and deeper, achieving CERs as low as 2.84% on a Japanese dataset with ensembling.
Conformer has achieved impressive results in Automatic Speech Recognition (ASR) by leveraging transformer's capturing of content-based global interactions and convolutional neural network's exploiting of local features. In Conformer, two macaron-like feed-forward layers with half-step residual connections sandwich the multi-head self-attention and convolution modules followed by a post layer normalization. We improve Conformer's long-sequence representation ability in two directions, \emph{sparser} and \emph{deeper}. We adapt a sparse self-attention mechanism with $\mathcal{O}(L\text{log}L)$ in time complexity and memory usage. A deep normalization strategy is utilized when performing residual connections to ensure our training of hundred-level Conformer blocks. On the Japanese CSJ-500h dataset, this deep sparse Conformer achieves respectively CERs of 5.52\%, 4.03\% and 4.50\% on the three evaluation sets and 4.16\%, 2.84\% and 3.20\% when ensembling five deep sparse Conformer variants from 12 to 16, 17, 50, and finally 100 encoder layers.