JHU IWSLT 2024 Dialectal and Low-resource System Description

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216208%3A11320%2F25%3APXAQ5NK5" target="_blank" >RIV/00216208:11320/25:PXAQ5NK5 - isvavai.cz</a>

Výsledek na webu

<a href="https://www.scopus.com/inward/record.uri?eid=2-s2.0-85204370217&partnerID=40&md5=805772f42861eabf1f8ee49cb9e8c57b" target="_blank" >https://www.scopus.com/inward/record.uri?eid=2-s2.0-85204370217&partnerID=40&md5=805772f42861eabf1f8ee49cb9e8c57b</a>

DOI - Digital Object Identifier

—

Jazyk výsledku

angličtina

Název v původním jazyce

JHU IWSLT 2024 Dialectal and Low-resource System Description

Popis výsledku v původním jazyce

Johns Hopkins University (JHU) submitted systems for all eight language pairs in the 2024 Low-Resource Language Track. The main effort of this work revolves around fine-tuning large and publicly available models in three proposed systems: i) end-to-end speech translation (ST) fine-tuning of SEAMLESSM4T v2; ii) ST fine-tuning of Whisper; iii) a cascaded system involving automatic speech recognition with fine-tuned Whisper and machine translation with NLLB. On top of systems above, we conduct a comparative analysis of different training paradigms, such as intra-distillation of NLLB, joint training and curriculum learning of SEAMLESSM4T v2, and multi-task learning and pseudo-translation with Whisper. Our results show that the best-performing approach differs by language pairs, but that i) fine-tuned SEAMLESSM4T v2 tends to perform best for source languages on which it was pre-trained, ii) multitask training helps Whisper fine-tuning, iii) cascaded systems with Whisper and NLLB tend to outperform Whisper alone, and iv) intra-distillation helps NLLB fine-tuning. ©2024 Association for Computational Linguistics.

Název v anglickém jazyce

JHU IWSLT 2024 Dialectal and Low-resource System Description

Popis výsledku anglicky

Johns Hopkins University (JHU) submitted systems for all eight language pairs in the 2024 Low-Resource Language Track. The main effort of this work revolves around fine-tuning large and publicly available models in three proposed systems: i) end-to-end speech translation (ST) fine-tuning of SEAMLESSM4T v2; ii) ST fine-tuning of Whisper; iii) a cascaded system involving automatic speech recognition with fine-tuned Whisper and machine translation with NLLB. On top of systems above, we conduct a comparative analysis of different training paradigms, such as intra-distillation of NLLB, joint training and curriculum learning of SEAMLESSM4T v2, and multi-task learning and pseudo-translation with Whisper. Our results show that the best-performing approach differs by language pairs, but that i) fine-tuned SEAMLESSM4T v2 tends to perform best for source languages on which it was pre-trained, ii) multitask training helps Whisper fine-tuning, iii) cascaded systems with Whisper and NLLB tend to outperform Whisper alone, and iv) intra-distillation helps NLLB fine-tuning. ©2024 Association for Computational Linguistics.

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

—

Rok uplatnění

2024

Kód důvěrnosti údajů

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

Název statě ve sborníku

IWSLT - Int. Conf. Spok. Lang. Transl., Proc. Conf.

ISBN

979-889176141-4

ISSN

—

e-ISSN

—

Počet stran výsledku

14

Strana od-do

188-201

Název nakladatele

Association for Computational Linguistics (ACL)

Místo vydání

—

Místo konání akce

Hybrid, Bangkok

Datum konání akce

1. 1. 2025

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

—