# Compute Cost Amortized Transformer for Streaming ASR

arXiv: https://arxiv.org/abs/2207.02393 · HTML: https://mlsanity.com/paper/2207.02393/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 23 of its 28 references that are in the map.
- First posted on arXiv 5 Jul 2022.

## Builds on (most cited first, 23 of 23)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://mlsanity.com/paper/1810.04805/) (2018)
- [Adam: A Method for Stochastic Optimization](https://mlsanity.com/paper/1412.6980/) (2014)
- [Categorical Reparameterization with Gumbel-Softmax](https://mlsanity.com/paper/1611.01144/) (2016)
- [Conformer: Convolution-augmented Transformer for Speech Recognition](https://mlsanity.com/paper/2005.08100/) (2020)
- [Big Bird: Transformers for Longer Sequences](https://mlsanity.com/paper/2007.14062/) (2020)
- [Reformer: The Efficient Transformer](https://mlsanity.com/paper/2001.04451/) (2020)
- [Rethinking Attention with Performers](https://mlsanity.com/paper/2009.14794/) (2020)
- [Linformer: Self-Attention with Linear Complexity](https://mlsanity.com/paper/2006.04768/) (2020)
- [Efficient Transformers: A Survey](https://mlsanity.com/paper/2009.06732/) (2020)
- [Speech Recognition with Deep Recurrent Neural Networks](https://mlsanity.com/paper/1303.5778/) (2013)
- [Sparse Sinkhorn Attention](https://mlsanity.com/paper/2002.11296/) (2020)
- [Transformer Transducer: A Streamable Speech Recognition Model with Transformer Encoders and RNN-T Loss](https://mlsanity.com/paper/2002.02562/) (2020)
- [A Comparative Study on Transformer vs RNN in Speech Applications](https://mlsanity.com/paper/1909.06317/) (2019)
- [Blockwise Self-Attention for Long Document Understanding](https://mlsanity.com/paper/1911.02972/) (2019)
- [Exploring Architectures, Data and Units For Streaming End-to-End Speech Recognition with RNN-Transducer](https://mlsanity.com/paper/1801.00841/) (2018)
- [Large-Scale Multilingual Speech Recognition with a Streaming End-to-End Model](https://mlsanity.com/paper/1909.05330/) (2019)
- [Transformer-Transducer: End-to-End Speech Recognition with Self-Attention](https://mlsanity.com/paper/1910.12977/) (2019)
- [Conv-Transformer Transducer: Low Latency, Low Frame Rate, Streamable End-to-End Speech Recognition](https://mlsanity.com/paper/2008.05750/) (2020)
- [Dynamic Encoder Transducer: A Flexible Solution For Trading Off Accuracy For Latency](https://mlsanity.com/paper/2104.02176/) (2021)
- [Amortized Neural Networks for Low-Latency Speech Recognition](https://mlsanity.com/paper/2108.01553/) (2021)
- [Bifocal Neural ASR: Exploiting Keyword Spotting for Inference Optimization](https://mlsanity.com/paper/2108.01704/) (2021)
- [Dual-Encoder Architecture with Encoder Selection for Joint Close-Talk and Far-Talk Speech Recognition](https://mlsanity.com/paper/2109.08744/) (2021)

## Built on by (most cited first, 2 of 2)

- [I3D: Transformer architectures with input-dependent dynamic depth for speech recognition](https://mlsanity.com/paper/2303.07624/) (2023)
- [Dual-Attention Neural Transducers for Efficient Wake Word Spotting in Speech Recognition](https://mlsanity.com/paper/2304.01905/) (2023)

## Abstract (arXiv)

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 computation pathways dynamically at inference time, resulting in selective use of compute resources throughout decoding, enabling significant reductions in compute with minimal impact on accuracy. The fully differentiable architecture is trained end-to-end with an accompanying lightweight arbitrator mechanism operating at the frame-level to make dynamic decisions on each input while a tunable loss function is used to regularize the overall level of compute against predictive performance. We report empirical results from experiments using the compute amortized Transformer-Transducer (T-T) model conducted on LibriSpeech data. Our best model can achieve a 60% compute cost reduction with only a 3% relative word error rate (WER) increase.