Espresso: A Fast End-to-end Neural Speech Recognition Toolkit
Yiming Wang, Tongfei Chen, Hainan Xu, Shuoyang Ding, Hang Lv, Yiwen Shao, Nanyun Peng, Lei Xie, Shinji Watanabe, Sanjeev Khudanpur
Code Available — Be the first to reproduce this paper.
ReproduceCode
- github.com/freewym/espressoOfficialIn paperpytorch★ 940
Abstract
We present Espresso, an open-source, modular, extensible end-to-end neural automatic speech recognition (ASR) toolkit based on the deep learning library PyTorch and the popular neural machine translation toolkit fairseq. Espresso supports distributed training across GPUs and computing nodes, and features various decoding approaches commonly employed in ASR, including look-ahead word-based language model fusion, for which a fast, parallelized decoder is implemented. Espresso achieves state-of-the-art ASR performance on the WSJ, LibriSpeech, and Switchboard data sets among other end-to-end systems without data augmentation, and is 4--11x faster for decoding than similar systems (e.g. ESPnet).
Tasks
Benchmark Results
| Dataset | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| Hub5'00 CallHome | Espresso | Word Error Rate (WER) | 19.1 | — | Unverified |
| Hub5'00 SwitchBoard | Espresso | Eval2000 | 9.2 | — | Unverified |
| LibriSpeech test-clean | Espresso | Word Error Rate (WER) | 2.8 | — | Unverified |
| LibriSpeech test-other | Espresso | Word Error Rate (WER) | 8.7 | — | Unverified |
| WSJ eval92 | Espresso | Word Error Rate (WER) | 3.4 | — | Unverified |