Parameter-Efficient Tuning With Adaptive Bottlenecks For Automatic Speech Recognition

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216305%3A26230%2F23%3APU150843" target="_blank" >RIV/00216305:26230/23:PU150843 - isvavai.cz</a>

Výsledek na webu

<a href="https://ieeexplore.ieee.org/document/10389769" target="_blank" >https://ieeexplore.ieee.org/document/10389769</a>

DOI - Digital Object Identifier

—

Jazyk výsledku

angličtina

Název v původním jazyce

Parameter-Efficient Tuning With Adaptive Bottlenecks For Automatic Speech Recognition

Popis výsledku v původním jazyce

Transfer learning from large multilingual pretrained models, like XLSR, has become the new paradigm for Automatic Speech Recognition (ASR). Considering their ever-increasing size, fine-tuning all the weights has become impractical when the computing budget is limited. Adapters are lightweight trainable modules inserted between layers while the pretrained part is kept frozen. They form a parameter-efficient fine-tuning method, but they still require a large bottleneck size to match standard fine-tuning performance. In this paper, we propose ABSADAPTER, a method to further reduce the parameter budget for equal task performance. Specifically, ABSADAPTER uses an Adaptive Bottleneck Scheduler to redistribute the adapter's weights to the layers that need adaptation the most. By training only 8% of the XLSR model, ABSADAPTER achieves close to standard fine-tuning performance on a domain-shifted Air-Traffic Communication (ATC) ASR task.

Název v anglickém jazyce

Parameter-Efficient Tuning With Adaptive Bottlenecks For Automatic Speech Recognition

Popis výsledku anglicky

Transfer learning from large multilingual pretrained models, like XLSR, has become the new paradigm for Automatic Speech Recognition (ASR). Considering their ever-increasing size, fine-tuning all the weights has become impractical when the computing budget is limited. Adapters are lightweight trainable modules inserted between layers while the pretrained part is kept frozen. They form a parameter-efficient fine-tuning method, but they still require a large bottleneck size to match standard fine-tuning performance. In this paper, we propose ABSADAPTER, a method to further reduce the parameter budget for equal task performance. Specifically, ABSADAPTER uses an Adaptive Bottleneck Scheduler to redistribute the adapter's weights to the layers that need adaptation the most. By training only 8% of the XLSR model, ABSADAPTER achieves close to standard fine-tuning performance on a domain-shifted Air-Traffic Communication (ATC) ASR task.

Druh

O - Ostatní výsledky

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