@inproceedings{waheed-etal-2024-distill,
title = "To Distill or Not to Distill? On the Robustness of Robust Knowledge Distillation",
author = "Waheed, Abdul and
Kadaoui, Karima and
Abdul-Mageed, Muhammad",
editor = "Ku, Lun-Wei and
Martins, Andre and
Srikumar, Vivek",
booktitle = "Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = aug,
year = "2024",
address = "Bangkok, Thailand",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.luhme-long.680/",
doi = "10.18653/v1/2024.acl-long.680",
pages = "12603--12621",
abstract = "Arabic is known to present unique challengesfor Automatic Speech Recognition (ASR). Onone hand, its rich linguistic diversity andwide range of dialects complicate the de-velopment of robust, inclusive models. Onthe other, current multilingual ASR modelsare compute-intensive and lack proper com-prehensive evaluations. In light of thesechallenges, we distill knowledge from largeteacher models into smaller student variantsthat more efficient. We also introduce a novelhuman-annotated dataset covering five under-represented Arabic dialects for evaluation. Wefurther evaluate both our models and existingSoTA multilingual models on both standardavailable benchmarks and our new dialectaldata. Our best-distilled model`s overall perfor-mance (45.0{\%} WER) surpasses that of a SoTAmodel twice its size (SeamlessM4T-large-v2,WER=47.0{\%}) and its teacher model (Whisper-large-v2, WER=55.1{\%}), and its average perfor-mance on our new dialectal data (56.9{\%} WER)outperforms all other models. To gain more in-sight into the poor performance of these modelson dialectal data, we conduct an error analysisand report the main types of errors the differentmodels tend to make. The GitHub repositoryfor the project is available at https://github.com/UBC-NLP/distill-whisper-ar."
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="waheed-etal-2024-distill">
<titleInfo>
<title>To Distill or Not to Distill? On the Robustness of Robust Knowledge Distillation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Abdul</namePart>
<namePart type="family">Waheed</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Karima</namePart>
<namePart type="family">Kadaoui</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Muhammad</namePart>
<namePart type="family">Abdul-Mageed</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2024-08</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Lun-Wei</namePart>
<namePart type="family">Ku</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Andre</namePart>
<namePart type="family">Martins</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Vivek</namePart>
<namePart type="family">Srikumar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Bangkok, Thailand</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Arabic is known to present unique challengesfor Automatic Speech Recognition (ASR). Onone hand, its rich linguistic diversity andwide range of dialects complicate the de-velopment of robust, inclusive models. Onthe other, current multilingual ASR modelsare compute-intensive and lack proper com-prehensive evaluations. In light of thesechallenges, we distill knowledge from largeteacher models into smaller student variantsthat more efficient. We also introduce a novelhuman-annotated dataset covering five under-represented Arabic dialects for evaluation. Wefurther evaluate both our models and existingSoTA multilingual models on both standardavailable benchmarks and our new dialectaldata. Our best-distilled model‘s overall perfor-mance (45.0% WER) surpasses that of a SoTAmodel twice its size (SeamlessM4T-large-v2,WER=47.0%) and its teacher model (Whisper-large-v2, WER=55.1%), and its average perfor-mance on our new dialectal data (56.9% WER)outperforms all other models. To gain more in-sight into the poor performance of these modelson dialectal data, we conduct an error analysisand report the main types of errors the differentmodels tend to make. The GitHub repositoryfor the project is available at https://github.com/UBC-NLP/distill-whisper-ar.</abstract>
<identifier type="citekey">waheed-etal-2024-distill</identifier>
<identifier type="doi">10.18653/v1/2024.acl-long.680</identifier>
<location>
<url>https://aclanthology.org/2024.luhme-long.680/</url>
</location>
<part>
<date>2024-08</date>
<extent unit="page">
<start>12603</start>
<end>12621</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T To Distill or Not to Distill? On the Robustness of Robust Knowledge Distillation
%A Waheed, Abdul
%A Kadaoui, Karima
%A Abdul-Mageed, Muhammad
%Y Ku, Lun-Wei
%Y Martins, Andre
%Y Srikumar, Vivek
%S Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
%D 2024
%8 August
%I Association for Computational Linguistics
%C Bangkok, Thailand
%F waheed-etal-2024-distill
%X Arabic is known to present unique challengesfor Automatic Speech Recognition (ASR). Onone hand, its rich linguistic diversity andwide range of dialects complicate the de-velopment of robust, inclusive models. Onthe other, current multilingual ASR modelsare compute-intensive and lack proper com-prehensive evaluations. In light of thesechallenges, we distill knowledge from largeteacher models into smaller student variantsthat more efficient. We also introduce a novelhuman-annotated dataset covering five under-represented Arabic dialects for evaluation. Wefurther evaluate both our models and existingSoTA multilingual models on both standardavailable benchmarks and our new dialectaldata. Our best-distilled model‘s overall perfor-mance (45.0% WER) surpasses that of a SoTAmodel twice its size (SeamlessM4T-large-v2,WER=47.0%) and its teacher model (Whisper-large-v2, WER=55.1%), and its average perfor-mance on our new dialectal data (56.9% WER)outperforms all other models. To gain more in-sight into the poor performance of these modelson dialectal data, we conduct an error analysisand report the main types of errors the differentmodels tend to make. The GitHub repositoryfor the project is available at https://github.com/UBC-NLP/distill-whisper-ar.
%R 10.18653/v1/2024.acl-long.680
%U https://aclanthology.org/2024.luhme-long.680/
%U https://doi.org/10.18653/v1/2024.acl-long.680
%P 12603-12621
Markdown (Informal)
[To Distill or Not to Distill? On the Robustness of Robust Knowledge Distillation](https://aclanthology.org/2024.luhme-long.680/) (Waheed et al., ACL 2024)
ACL