@inproceedings{quteineh-etal-2022-enhancing,
title = "Enhancing Task-Specific Distillation in Small Data Regimes through Language Generation",
author = "Quteineh, Husam and
Samothrakis, Spyridon and
Sutcliffe, Richard",
editor = "Calzolari, Nicoletta and
Huang, Chu-Ren and
Kim, Hansaem and
Pustejovsky, James and
Wanner, Leo and
Choi, Key-Sun and
Ryu, Pum-Mo and
Chen, Hsin-Hsi and
Donatelli, Lucia and
Ji, Heng and
Kurohashi, Sadao and
Paggio, Patrizia and
Xue, Nianwen and
Kim, Seokhwan and
Hahm, Younggyun and
He, Zhong and
Lee, Tony Kyungil and
Santus, Enrico and
Bond, Francis and
Na, Seung-Hoon",
booktitle = "Proceedings of the 29th International Conference on Computational Linguistics",
month = oct,
year = "2022",
address = "Gyeongju, Republic of Korea",
publisher = "International Committee on Computational Linguistics",
url = "https://aclanthology.org/2022.coling-1.520",
pages = "5955--5965",
abstract = "Large-scale pretrained language models have led to significant improvements in Natural Language Processing. Unfortunately, they come at the cost of high computational and storage requirements that complicate their deployment on low-resource devices. This issue can be addressed by distilling knowledge from larger models to smaller ones through pseudo-labels on task-specific datasets. However, this can be difficult for tasks with very limited data. To overcome this challenge, we present a novel approach where knowledge can be distilled from a teacher model to a student model through the generation of synthetic data. For this to be done, we first fine-tune the teacher and student models, as well as a Natural Language Generation (NLG) model, on the target task dataset. We then let both student and teacher work together to condition the NLG model to generate examples that can enhance the performance of the student. We tested our approach on two data generation methods: a) Targeted generation using the Monte Carlo Tree Search (MCTS) algorithm, and b) A Non-Targeted Text Generation (NTTG) method. We evaluate the effectiveness of our approaches against a baseline that uses the BERT model for data augmentation through random word replacement. By testing this approach on the SST-2, MRPC, YELP-2, DBpedia, and TREC-6 datasets, we consistently witnessed considerable improvements over the word-replacement baseline.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="quteineh-etal-2022-enhancing">
<titleInfo>
<title>Enhancing Task-Specific Distillation in Small Data Regimes through Language Generation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Husam</namePart>
<namePart type="family">Quteineh</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Spyridon</namePart>
<namePart type="family">Samothrakis</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Richard</namePart>
<namePart type="family">Sutcliffe</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2022-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 29th International Conference on Computational Linguistics</title>
</titleInfo>
<name type="personal">
<namePart type="given">Nicoletta</namePart>
<namePart type="family">Calzolari</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Chu-Ren</namePart>
<namePart type="family">Huang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hansaem</namePart>
<namePart type="family">Kim</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">James</namePart>
<namePart type="family">Pustejovsky</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Leo</namePart>
<namePart type="family">Wanner</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Key-Sun</namePart>
<namePart type="family">Choi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Pum-Mo</namePart>
<namePart type="family">Ryu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hsin-Hsi</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lucia</namePart>
<namePart type="family">Donatelli</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Heng</namePart>
<namePart type="family">Ji</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sadao</namePart>
<namePart type="family">Kurohashi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Patrizia</namePart>
<namePart type="family">Paggio</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nianwen</namePart>
<namePart type="family">Xue</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Seokhwan</namePart>
<namePart type="family">Kim</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Younggyun</namePart>
<namePart type="family">Hahm</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhong</namePart>
<namePart type="family">He</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tony</namePart>
<namePart type="given">Kyungil</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Enrico</namePart>
<namePart type="family">Santus</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Francis</namePart>
<namePart type="family">Bond</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Seung-Hoon</namePart>
<namePart type="family">Na</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>International Committee on Computational Linguistics</publisher>
<place>
<placeTerm type="text">Gyeongju, Republic of Korea</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Large-scale pretrained language models have led to significant improvements in Natural Language Processing. Unfortunately, they come at the cost of high computational and storage requirements that complicate their deployment on low-resource devices. This issue can be addressed by distilling knowledge from larger models to smaller ones through pseudo-labels on task-specific datasets. However, this can be difficult for tasks with very limited data. To overcome this challenge, we present a novel approach where knowledge can be distilled from a teacher model to a student model through the generation of synthetic data. For this to be done, we first fine-tune the teacher and student models, as well as a Natural Language Generation (NLG) model, on the target task dataset. We then let both student and teacher work together to condition the NLG model to generate examples that can enhance the performance of the student. We tested our approach on two data generation methods: a) Targeted generation using the Monte Carlo Tree Search (MCTS) algorithm, and b) A Non-Targeted Text Generation (NTTG) method. We evaluate the effectiveness of our approaches against a baseline that uses the BERT model for data augmentation through random word replacement. By testing this approach on the SST-2, MRPC, YELP-2, DBpedia, and TREC-6 datasets, we consistently witnessed considerable improvements over the word-replacement baseline.</abstract>
<identifier type="citekey">quteineh-etal-2022-enhancing</identifier>
<location>
<url>https://aclanthology.org/2022.coling-1.520</url>
</location>
<part>
<date>2022-10</date>
<extent unit="page">
<start>5955</start>
<end>5965</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Enhancing Task-Specific Distillation in Small Data Regimes through Language Generation
%A Quteineh, Husam
%A Samothrakis, Spyridon
%A Sutcliffe, Richard
%Y Calzolari, Nicoletta
%Y Huang, Chu-Ren
%Y Kim, Hansaem
%Y Pustejovsky, James
%Y Wanner, Leo
%Y Choi, Key-Sun
%Y Ryu, Pum-Mo
%Y Chen, Hsin-Hsi
%Y Donatelli, Lucia
%Y Ji, Heng
%Y Kurohashi, Sadao
%Y Paggio, Patrizia
%Y Xue, Nianwen
%Y Kim, Seokhwan
%Y Hahm, Younggyun
%Y He, Zhong
%Y Lee, Tony Kyungil
%Y Santus, Enrico
%Y Bond, Francis
%Y Na, Seung-Hoon
%S Proceedings of the 29th International Conference on Computational Linguistics
%D 2022
%8 October
%I International Committee on Computational Linguistics
%C Gyeongju, Republic of Korea
%F quteineh-etal-2022-enhancing
%X Large-scale pretrained language models have led to significant improvements in Natural Language Processing. Unfortunately, they come at the cost of high computational and storage requirements that complicate their deployment on low-resource devices. This issue can be addressed by distilling knowledge from larger models to smaller ones through pseudo-labels on task-specific datasets. However, this can be difficult for tasks with very limited data. To overcome this challenge, we present a novel approach where knowledge can be distilled from a teacher model to a student model through the generation of synthetic data. For this to be done, we first fine-tune the teacher and student models, as well as a Natural Language Generation (NLG) model, on the target task dataset. We then let both student and teacher work together to condition the NLG model to generate examples that can enhance the performance of the student. We tested our approach on two data generation methods: a) Targeted generation using the Monte Carlo Tree Search (MCTS) algorithm, and b) A Non-Targeted Text Generation (NTTG) method. We evaluate the effectiveness of our approaches against a baseline that uses the BERT model for data augmentation through random word replacement. By testing this approach on the SST-2, MRPC, YELP-2, DBpedia, and TREC-6 datasets, we consistently witnessed considerable improvements over the word-replacement baseline.
%U https://aclanthology.org/2022.coling-1.520
%P 5955-5965
Markdown (Informal)
[Enhancing Task-Specific Distillation in Small Data Regimes through Language Generation](https://aclanthology.org/2022.coling-1.520) (Quteineh et al., COLING 2022)
ACL