HyperConformer: Multi-head HyperMixer for Efficient Speech Recognition

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216305%3A26230%2F23%3APU150721" target="_blank" >RIV/00216305:26230/23:PU150721 - isvavai.cz</a>

Výsledek na webu

<a href="https://www.isca-archive.org/interspeech_2023/mai23_interspeech.pdf" target="_blank" >https://www.isca-archive.org/interspeech_2023/mai23_interspeech.pdf</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.21437/Interspeech.2023-1611" target="_blank" >10.21437/Interspeech.2023-1611</a>

Jazyk výsledku

angličtina

Název v původním jazyce

HyperConformer: Multi-head HyperMixer for Efficient Speech Recognition

Popis výsledku v původním jazyce

State-of-the-art ASR systems have achieved promising results by modeling local and global interactions separately. While the former can be computed efficiently, global interactions are usu- ally modeled via attention mechanisms, which are expensive for long input sequences. Here, we address this by extending Hy- perMixer, an efficient alternative to attention exhibiting linear complexity, to the Conformer architecture for speech recogni- tion, leading to HyperConformer. In particular, multi-head Hy- perConformer achieves comparable or higher recognition per- formance while being more efficient than Conformer in terms of inference speed, memory, parameter count, and available train- ing data. HyperConformer achieves a word error rate of 2.9% on LibriSpeech test-clean with less than 8M neural parameters and a peak memory during training of 5.7GB, hence trainable with accessible hardware. Encoder speed is between 38% on mid-length speech and 56% on long speech faster than an equiv- alent Conformer.1)

Název v anglickém jazyce

HyperConformer: Multi-head HyperMixer for Efficient Speech Recognition

Popis výsledku anglicky

State-of-the-art ASR systems have achieved promising results by modeling local and global interactions separately. While the former can be computed efficiently, global interactions are usu- ally modeled via attention mechanisms, which are expensive for long input sequences. Here, we address this by extending Hy- perMixer, an efficient alternative to attention exhibiting linear complexity, to the Conformer architecture for speech recogni- tion, leading to HyperConformer. In particular, multi-head Hy- perConformer achieves comparable or higher recognition per- formance while being more efficient than Conformer in terms of inference speed, memory, parameter count, and available train- ing data. HyperConformer achieves a word error rate of 2.9% on LibriSpeech test-clean with less than 8M neural parameters and a peak memory during training of 5.7GB, hence trainable with accessible hardware. Encoder speed is between 38% on mid-length speech and 56% on long speech faster than an equiv- alent Conformer.1)

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Projekt

—

Návaznosti

S - Specificky vyzkum na vysokych skolach<br>I - Institucionalni podpora na dlouhodoby koncepcni rozvoj vyzkumne organizace

Rok uplatnění

2023

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název statě ve sborníku

Proceedings of the Annual Conference of International Speech Communication Association, INTERSPEECH

ISBN

—

ISSN

1990-9772

e-ISSN

—

Počet stran výsledku

5

Strana od-do

2213-2217

Název nakladatele

International Speech Communication Association

Místo vydání

Dublin

Místo konání akce

Dublin

Datum konání akce

20. 8. 2023

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

—