paper-with-me

홈 › Papers

Amortized Neural Networks for Low-Latency Speech Recognition

2021-08-03 · Jonathan Macoskey, Grant P. Strimel, Jinru Su, Ariya Rastrow

We introduce Amortized Neural Networks (AmNets), a compute cost- and latency-aware network architecture particularly well-suited for sequence modeling tasks. We apply AmNets to the Recurrent Neural Network Transducer (RNN-T) to reduce compute cost and latency for an automatic speech recognition (ASR) task. The AmNets RNN-T architecture enables the network to dynamically switch between encoder branches on a frame-by-frame basis. Branches are constructed with variable levels of compute cost and model capacity. Here, we achieve variable compute for two well-known candidate techniques: one using sparse pruning and the other using matrix factorization. Frame-by-frame switching is determined by an arbitrator network that requires negligible compute overhead. We present results using both architectures on LibriSpeech data and show that our proposed architecture can reduce inference cost by up to 45\% and latency to nearly real-time without incurring a loss in accuracy.

📄 PDF Abstract BibTeX arXiv:2108.01553

Code (0)

등록된 구현이 없습니다.

Tasks

Automatic Speech RecognitionAutomatic Speech Recognition (ASR)speech-recognitionSpeech Recognition

Methods 이 논문이 사용한 방법론

Pruning 설명 없음

Similar Papers 제목 키워드 기반

Compute Cost Amortized Transformer for Streaming ASR

2022-07-05 · Yi Xie, Jonathan Macoskey, Martin Radfar, Feng-Ju Chang 외

We present a streaming, Transformer-based end-to-end automatic speech recognition (ASR) architecture which achieves efficient neural inference through compute cost amortization. Our architecture creates sparse computatio…

Automatic Speech RecognitionAutomatic Speech Recognition (ASR)speech-recognitionSpeech Recognition

Transformer Transducer: One Model Unifying Streaming and Non-streaming Speech Recognition

2020-10-07 · Anshuman Tripathi, Jaeyoung Kim, Qian Zhang, Han Lu 외

In this paper we present a Transformer-Transducer model architecture and a training technique to unify streaming and non-streaming speech recognition models into one model. The model is composed of a stack of transformer…

speech-recognitionSpeech Recognition

Dynamic latency speech recognition with asynchronous revision

2020-11-03 · Mingkun Huang, Meng Cai, Jun Zhang, Yang Zhang 외

In this work we propose an inference technique, asynchronous revision, to unify streaming and non-streaming speech recognition models. Specifically, we achieve dynamic latency with only one model by using arbitrary right…

Decoderspeech-recognitionSpeech Recognition

Dynamic Latency for CTC-Based Streaming Automatic Speech Recognition With Emformer

2022-03-29 · Jingyu Sun, Guiping Zhong, Dinghao Zhou, Baoxiang Li

An inferior performance of the streaming automatic speech recognition models versus non-streaming model is frequently seen due to the absence of future context. In order to improve the performance of the streaming model …

Automatic Speech RecognitionAutomatic Speech Recognition (ASR)speech-recognitionSpeech Recognition

Dissecting User-Perceived Latency of On-Device E2E Speech Recognition

2021-04-06 · Yuan Shangguan, Rohit Prabhavalkar, Hang Su, Jay Mahadeokar 외

As speech-enabled devices such as smartphones and smart speakers become increasingly ubiquitous, there is growing interest in building automatic speech recognition (ASR) systems that can run directly on-device; end-to-en…

Automatic Speech RecognitionAutomatic Speech Recognition (ASR)speech-recognitionSpeech Recognition