Jasper: An End-to-End Convolutional Neural Acoustic Model

Code

• github.com/osmr/imgclsmob
  mxnet★ 3,015
• github.com/TensorSpeech/TensorFlowASR
  tf★ 1,005
• github.com/msalhab96/SpeeQ
  pytorch★ 51
• github.com/sooftware/OpenSpeech
  pytorch★ 35
• github.com/NVIDIA/OpenSeq2Seq
  tf★ 0
• github.com/stefanpantic/asr
  tf★ 0
• github.com/MindSpore-paper-code-2/code399/tree/main/jasper
  mindspore★ 0
• github.com/2023-MindSpore-1/ms-code-15/tree/main/jasper
  mindspore★ 0
• github.com/sooftware/jasper-pytorch
  pytorch★ 0
• github.com/marka17/digit-recognition
  pytorch★ 0

Abstract

In this paper, we report state-of-the-art results on LibriSpeech among end-to-end speech recognition models without any external training data. Our model, Jasper, uses only 1D convolutions, batch normalization, ReLU, dropout, and residual connections. To improve training, we further introduce a new layer-wise optimizer called NovoGrad. Through experiments, we demonstrate that the proposed deep architecture performs as well or better than more complex choices. Our deepest Jasper variant uses 54 convolutional layers. With this architecture, we achieve 2.95% WER using a beam-search decoder with an external neural language model and 3.86% WER with a greedy decoder on LibriSpeech test-clean. We also report competitive results on the Wall Street Journal and the Hub5'00 conversational evaluation datasets.

Tasks

Benchmark Results

Reproductions