Improving Speaker Verification with Self-Pretrained Transformer Models

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216305%3A26230%2F23%3APU149718" target="_blank" >RIV/00216305:26230/23:PU149718 - isvavai.cz</a>

Výsledek na webu

<a href="https://www.isca-speech.org/archive/pdfs/interspeech_2023/peng23_interspeech.pdf" target="_blank" >https://www.isca-speech.org/archive/pdfs/interspeech_2023/peng23_interspeech.pdf</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.21437/Interspeech.2023-453" target="_blank" >10.21437/Interspeech.2023-453</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Improving Speaker Verification with Self-Pretrained Transformer Models

Popis výsledku v původním jazyce

Recently, fine-tuning large pre-trained Transformer models using downstream datasets has received a rising interest. Despite their success, it is still challenging to disentangle the benefits of large-scale datasets and Transformer structures from the limitations of the pre-training. In this paper, we introduce a hierarchical training approach, named self-pretraining, in which Transformer models are pretrained and finetuned on the same dataset. Three pre-trained models including HuBERT, Conformer andWavLM are evaluated on four different speaker verification datasets with varying sizes. Our experiments show that these self-pretrained models achieve competitive performance on downstream speaker verification tasks with only one-third of the data compared to Librispeech pretraining, such as Vox- Celeb1 and CNCeleb1. Furthermore, when pre-training only on the VoxCeleb2-dev, the Conformer model outperforms the one pre-trained on 94k hours of data using the same fine-tuning settings.

Název v anglickém jazyce

Improving Speaker Verification with Self-Pretrained Transformer Models

Popis výsledku anglicky

Recently, fine-tuning large pre-trained Transformer models using downstream datasets has received a rising interest. Despite their success, it is still challenging to disentangle the benefits of large-scale datasets and Transformer structures from the limitations of the pre-training. In this paper, we introduce a hierarchical training approach, named self-pretraining, in which Transformer models are pretrained and finetuned on the same dataset. Three pre-trained models including HuBERT, Conformer andWavLM are evaluated on four different speaker verification datasets with varying sizes. Our experiments show that these self-pretrained models achieve competitive performance on downstream speaker verification tasks with only one-third of the data compared to Librispeech pretraining, such as Vox- Celeb1 and CNCeleb1. Furthermore, when pre-training only on the VoxCeleb2-dev, the Conformer model outperforms the one pre-trained on 94k hours of data using the same fine-tuning settings.

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

Výsledek vznikl pri realizaci vícero projektů. Více informací v záložce Projekty.

Návaznosti

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)

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 the International Speech Communication Association, INTERSPEECH

ISBN

—

ISSN

1990-9772

e-ISSN

—

Počet stran výsledku

5

Strana od-do

5361-5365

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

—