How transfer learning impacts linguistic knowledge in deep NLP models?

Transfer learning from pre-trained neural language models towards downstream tasks has been a predominant theme in NLP recently. Several researchers have shown that deep NLP models learn non-trivial amount of linguistic knowledge, captured at different layers of the model. We investigate how fine-tu...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Durrani, Nadir, Sajjad, Hassan, Dalvi, Fahim
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page
container_issue
container_start_page
container_title
container_volume
creator Durrani, Nadir
Sajjad, Hassan
Dalvi, Fahim
description Transfer learning from pre-trained neural language models towards downstream tasks has been a predominant theme in NLP recently. Several researchers have shown that deep NLP models learn non-trivial amount of linguistic knowledge, captured at different layers of the model. We investigate how fine-tuning towards downstream NLP tasks impacts the learned linguistic knowledge. We carry out a study across popular pre-trained models BERT, RoBERTa and XLNet using layer and neuron-level diagnostic classifiers. We found that for some GLUE tasks, the network relies on the core linguistic information and preserve it deeper in the network, while for others it forgets. Linguistic information is distributed in the pre-trained language models but becomes localized to the lower layers post fine-tuning, reserving higher layers for the task specific knowledge. The pattern varies across architectures, with BERT retaining linguistic information relatively deeper in the network compared to RoBERTa and XLNet, where it is predominantly delegated to the lower layers.
doi_str_mv 10.48550/arxiv.2105.15179
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2105_15179</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2105_15179</sourcerecordid><originalsourceid>FETCH-LOGICAL-a679-c553af6f1b1f844c1dd4c256299c9ef56ab9c55fdaa86e7e50ec6ac1abd890f83</originalsourceid><addsrcrecordid>eNotz7FOwzAYBGAvDKjwAEz4BRLsJHbsqUIVpUgRMHSP_ti_K6uOE9mBwttTCtPppNNJHyF3nJWNEoI9QPryn2XFmSi54K2-JtvddKJLgpgdJhoQUvTxQP04g1kyDefy4fPiDT3G6RTQHpD6SC3iTF-7dzpOFkNe35ArByHj7X-uyH77tN_siu7t-WXz2BUgW10YIWpw0vGBO9U0hlvbmErISmuj0QkJgz5vnAVQElsUDI0Ew2GwSjOn6hW5_7u9QPo5-RHSd_8L6i-g-gdQcEcU</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>How transfer learning impacts linguistic knowledge in deep NLP models?</title><source>arXiv.org</source><creator>Durrani, Nadir ; Sajjad, Hassan ; Dalvi, Fahim</creator><creatorcontrib>Durrani, Nadir ; Sajjad, Hassan ; Dalvi, Fahim</creatorcontrib><description>Transfer learning from pre-trained neural language models towards downstream tasks has been a predominant theme in NLP recently. Several researchers have shown that deep NLP models learn non-trivial amount of linguistic knowledge, captured at different layers of the model. We investigate how fine-tuning towards downstream NLP tasks impacts the learned linguistic knowledge. We carry out a study across popular pre-trained models BERT, RoBERTa and XLNet using layer and neuron-level diagnostic classifiers. We found that for some GLUE tasks, the network relies on the core linguistic information and preserve it deeper in the network, while for others it forgets. Linguistic information is distributed in the pre-trained language models but becomes localized to the lower layers post fine-tuning, reserving higher layers for the task specific knowledge. The pattern varies across architectures, with BERT retaining linguistic information relatively deeper in the network compared to RoBERTa and XLNet, where it is predominantly delegated to the lower layers.</description><identifier>DOI: 10.48550/arxiv.2105.15179</identifier><language>eng</language><subject>Computer Science - Computation and Language</subject><creationdate>2021-05</creationdate><rights>http://arxiv.org/licenses/nonexclusive-distrib/1.0</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,780,885</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/2105.15179$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2105.15179$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Durrani, Nadir</creatorcontrib><creatorcontrib>Sajjad, Hassan</creatorcontrib><creatorcontrib>Dalvi, Fahim</creatorcontrib><title>How transfer learning impacts linguistic knowledge in deep NLP models?</title><description>Transfer learning from pre-trained neural language models towards downstream tasks has been a predominant theme in NLP recently. Several researchers have shown that deep NLP models learn non-trivial amount of linguistic knowledge, captured at different layers of the model. We investigate how fine-tuning towards downstream NLP tasks impacts the learned linguistic knowledge. We carry out a study across popular pre-trained models BERT, RoBERTa and XLNet using layer and neuron-level diagnostic classifiers. We found that for some GLUE tasks, the network relies on the core linguistic information and preserve it deeper in the network, while for others it forgets. Linguistic information is distributed in the pre-trained language models but becomes localized to the lower layers post fine-tuning, reserving higher layers for the task specific knowledge. The pattern varies across architectures, with BERT retaining linguistic information relatively deeper in the network compared to RoBERTa and XLNet, where it is predominantly delegated to the lower layers.</description><subject>Computer Science - Computation and Language</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2021</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotz7FOwzAYBGAvDKjwAEz4BRLsJHbsqUIVpUgRMHSP_ti_K6uOE9mBwttTCtPppNNJHyF3nJWNEoI9QPryn2XFmSi54K2-JtvddKJLgpgdJhoQUvTxQP04g1kyDefy4fPiDT3G6RTQHpD6SC3iTF-7dzpOFkNe35ArByHj7X-uyH77tN_siu7t-WXz2BUgW10YIWpw0vGBO9U0hlvbmErISmuj0QkJgz5vnAVQElsUDI0Ew2GwSjOn6hW5_7u9QPo5-RHSd_8L6i-g-gdQcEcU</recordid><startdate>20210531</startdate><enddate>20210531</enddate><creator>Durrani, Nadir</creator><creator>Sajjad, Hassan</creator><creator>Dalvi, Fahim</creator><scope>AKY</scope><scope>GOX</scope></search><sort><creationdate>20210531</creationdate><title>How transfer learning impacts linguistic knowledge in deep NLP models?</title><author>Durrani, Nadir ; Sajjad, Hassan ; Dalvi, Fahim</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a679-c553af6f1b1f844c1dd4c256299c9ef56ab9c55fdaa86e7e50ec6ac1abd890f83</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2021</creationdate><topic>Computer Science - Computation and Language</topic><toplevel>online_resources</toplevel><creatorcontrib>Durrani, Nadir</creatorcontrib><creatorcontrib>Sajjad, Hassan</creatorcontrib><creatorcontrib>Dalvi, Fahim</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Durrani, Nadir</au><au>Sajjad, Hassan</au><au>Dalvi, Fahim</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>How transfer learning impacts linguistic knowledge in deep NLP models?</atitle><date>2021-05-31</date><risdate>2021</risdate><abstract>Transfer learning from pre-trained neural language models towards downstream tasks has been a predominant theme in NLP recently. Several researchers have shown that deep NLP models learn non-trivial amount of linguistic knowledge, captured at different layers of the model. We investigate how fine-tuning towards downstream NLP tasks impacts the learned linguistic knowledge. We carry out a study across popular pre-trained models BERT, RoBERTa and XLNet using layer and neuron-level diagnostic classifiers. We found that for some GLUE tasks, the network relies on the core linguistic information and preserve it deeper in the network, while for others it forgets. Linguistic information is distributed in the pre-trained language models but becomes localized to the lower layers post fine-tuning, reserving higher layers for the task specific knowledge. The pattern varies across architectures, with BERT retaining linguistic information relatively deeper in the network compared to RoBERTa and XLNet, where it is predominantly delegated to the lower layers.</abstract><doi>10.48550/arxiv.2105.15179</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.2105.15179
ispartof
issn
language eng
recordid cdi_arxiv_primary_2105_15179
source arXiv.org
subjects Computer Science - Computation and Language
title How transfer learning impacts linguistic knowledge in deep NLP models?
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-29T18%3A25%3A40IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-arxiv_GOX&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=How%20transfer%20learning%20impacts%20linguistic%20knowledge%20in%20deep%20NLP%20models?&rft.au=Durrani,%20Nadir&rft.date=2021-05-31&rft_id=info:doi/10.48550/arxiv.2105.15179&rft_dat=%3Carxiv_GOX%3E2105_15179%3C/arxiv_GOX%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_id=info:pmid/&rfr_iscdi=true