Publicação

Open-Source Language Models for News Classification: Implementing Small Models in Low-Resource Environments

Ver documento

Detalhes bibliográficos
Resumo:Pre-Trained Models (PTMs) or Large Language Models (LLMs) are deep neural networks trained on vast amounts of text data, enabling them to make predictions based on learned knowledge. Google has played a significant role in this field, particularly through popularizing the Transformers Architecture. However, the landscape evolved dramatically with the release of ChatGPT by OpenAI in November 2022, marking the advent of the universal artificial intelligence era. This event sparked significant interest and efforts in studying LLMs, prompting industries to adapt their operations, software providers to refine their skills, and society to contemplate ethical implications. This research delves into the use of Open-Source LLMs, focusing particularly on text classification — a critical task in Natural Language Processing (NLP). The study employs techniques such as fine-tuning and model quantization, which are essential for leveraging LLMs effectively in practical applications. Key questions addressed include evaluating the comparability of open-source models with established benchmarks across different text classification approaches. The research aims to identify primary challenges and limitations associated with running modern open-source LLMs in lowresource environments. By exploring these topics, the research aims to contribute insights into optimizing the deployment of open-source LLMs, enhancing their accessibility, and addressing practical constraints that affect their widespread adoption across various sectors and applications in NLP.
Autores principais:Figueiredo, Fernando Niglio de
Assunto:Large Language Models (LLM) Pre-trained models (PTM) Text classification News classification Fine-tuning Open Source LoRA QLoRA low-resource environment SDG 8 - Decent work and economic growth
Ano:2024
País:Portugal
Tipo de documento:dissertação de mestrado
Tipo de acesso:acesso aberto
Instituição associada:Universidade Nova de Lisboa
Idioma:inglês
Origem:Repositório Institucional da UNL
_version_ 1865920617045295104
author Figueiredo, Fernando Niglio de
author_facet Figueiredo, Fernando Niglio de
Figueiredo, Fernando Niglio de
author_role author
contributor_name_str_mv Bação, Fernando José Ferreira Lucas
RUN
country_str PT
creators_json_str [{\"Person.name\":\"Figueiredo, Fernando Niglio de\"}]
datacite.contributors.contributor.contributorName.fl_str_mv Bação, Fernando José Ferreira Lucas
RUN
datacite.creators.creator.creatorName.fl_str_mv Figueiredo, Fernando Niglio de
datacite.date.Accepted.fl_str_mv 2024-10-24T00:00:00Z
datacite.date.available.fl_str_mv 2024-11-26T15:26:08Z
datacite.date.embargoed.fl_str_mv 2024-11-26T15:26:08Z
datacite.rights.fl_str_mv http://purl.org/coar/access_right/c_abf2
datacite.subjects.subject.fl_str_mv Large Language Models (LLM)
Pre-trained models (PTM)
Text classification
News classification
Fine-tuning
Open Source
LoRA
QLoRA
low-resource environment
SDG 8 - Decent work and economic growth
datacite.titles.title.fl_str_mv Open-Source Language Models for News Classification: Implementing Small Models in Low-Resource Environments
dc.contributor.none.fl_str_mv Bação, Fernando José Ferreira Lucas
RUN
dc.creator.none.fl_str_mv Figueiredo, Fernando Niglio de
dc.date.Accepted.fl_str_mv 2024-10-24T00:00:00Z
dc.date.available.fl_str_mv 2024-11-26T15:26:08Z
dc.date.embargoed.fl_str_mv 2024-11-26T15:26:08Z
dc.format.none.fl_str_mv application/pdf
dc.identifier.none.fl_str_mv http://hdl.handle.net/10362/175829
dc.language.none.fl_str_mv eng
dc.rights.cclincense.fl_str_mv http://creativecommons.org/licenses/by/4.0/
dc.rights.none.fl_str_mv http://purl.org/coar/access_right/c_abf2
dc.subject.none.fl_str_mv Large Language Models (LLM)
Pre-trained models (PTM)
Text classification
News classification
Fine-tuning
Open Source
LoRA
QLoRA
low-resource environment
SDG 8 - Decent work and economic growth
dc.title.fl_str_mv Open-Source Language Models for News Classification: Implementing Small Models in Low-Resource Environments
dc.type.none.fl_str_mv http://purl.org/coar/resource_type/c_bdcc
description Pre-Trained Models (PTMs) or Large Language Models (LLMs) are deep neural networks trained on vast amounts of text data, enabling them to make predictions based on learned knowledge. Google has played a significant role in this field, particularly through popularizing the Transformers Architecture. However, the landscape evolved dramatically with the release of ChatGPT by OpenAI in November 2022, marking the advent of the universal artificial intelligence era. This event sparked significant interest and efforts in studying LLMs, prompting industries to adapt their operations, software providers to refine their skills, and society to contemplate ethical implications. This research delves into the use of Open-Source LLMs, focusing particularly on text classification — a critical task in Natural Language Processing (NLP). The study employs techniques such as fine-tuning and model quantization, which are essential for leveraging LLMs effectively in practical applications. Key questions addressed include evaluating the comparability of open-source models with established benchmarks across different text classification approaches. The research aims to identify primary challenges and limitations associated with running modern open-source LLMs in lowresource environments. By exploring these topics, the research aims to contribute insights into optimizing the deployment of open-source LLMs, enhancing their accessibility, and addressing practical constraints that affect their widespread adoption across various sectors and applications in NLP.
dirty 0
eu_rights_str_mv openAccess
format masterThesis
fulltext.url.fl_str_mv https://run.unl.pt/bitstreams/f8b0e29f-5432-4d89-99a1-0de41082c908/download
id run_7b14e00a5107a54ca6c569dbfcd00dbf
identifier.url.fl_str_mv http://hdl.handle.net/10362/175829
instacron_str unl
institution Universidade Nova de Lisboa
instname_str Universidade Nova de Lisboa
language eng
network_acronym_str run
network_name_str Repositório Institucional da UNL
oai_identifier_str oai:run.unl.pt:10362/175829
organization_str_mv urn:organizationAcronym:unl
person_str_mv Figueiredo, Fernando Niglio de
publishDate 2024
reponame_str Repositório Institucional da UNL
repository_id_str urn:repositoryAcronym:run
service_str_mv urn:repositoryAcronym:run
spelling engpt_PTPre-Trained Models (PTMs) or Large Language Models (LLMs) are deep neural networks trained on vast amounts of text data, enabling them to make predictions based on learned knowledge. Google has played a significant role in this field, particularly through popularizing the Transformers Architecture. However, the landscape evolved dramatically with the release of ChatGPT by OpenAI in November 2022, marking the advent of the universal artificial intelligence era. This event sparked significant interest and efforts in studying LLMs, prompting industries to adapt their operations, software providers to refine their skills, and society to contemplate ethical implications. This research delves into the use of Open-Source LLMs, focusing particularly on text classification — a critical task in Natural Language Processing (NLP). The study employs techniques such as fine-tuning and model quantization, which are essential for leveraging LLMs effectively in practical applications. Key questions addressed include evaluating the comparability of open-source models with established benchmarks across different text classification approaches. The research aims to identify primary challenges and limitations associated with running modern open-source LLMs in lowresource environments. By exploring these topics, the research aims to contribute insights into optimizing the deployment of open-source LLMs, enhancing their accessibility, and addressing practical constraints that affect their widespread adoption across various sectors and applications in NLP.application/pdfpt_PTOpen-Source Language Models for News Classification: Implementing Small Models in Low-Resource EnvironmentsFigueiredo, Fernando Niglio deBação, Fernando José Ferreira LucasHostingInstitutionOrganizationalRUNe-mailmailto:run@unl.ptrun@unl.ptURNurn:tid:2037851422024-11-26T15:26:08Z2024-10-242024-10-24T00:00:00ZHandlehttp://hdl.handle.net/10362/175829http://purl.org/coar/access_right/c_abf2open accessLarge Language Models (LLM)Pre-trained models (PTM)Text classificationNews classificationFine-tuningOpen SourceLoRAQLoRAlow-resource environmentSDG 8 - Decent work and economic growth1432890 bytesliteraturehttp://purl.org/coar/resource_type/c_bdccmaster thesis2024-10-24http://creativecommons.org/licenses/by/4.0/http://purl.org/coar/access_right/c_abf2application/pdffulltexthttps://run.unl.pt/bitstreams/f8b0e29f-5432-4d89-99a1-0de41082c908/download
spellingShingle Open-Source Language Models for News Classification: Implementing Small Models in Low-Resource Environments
Open-Source Language Models for News Classification: Implementing Small Models in Low-Resource Environments
Figueiredo, Fernando Niglio de
Large Language Models (LLM)
Pre-trained models (PTM)
Text classification
News classification
Fine-tuning
Open Source
LoRA
QLoRA
low-resource environment
SDG 8 - Decent work and economic growth
Figueiredo, Fernando Niglio de
Large Language Models (LLM)
Pre-trained models (PTM)
Text classification
News classification
Fine-tuning
Open Source
LoRA
QLoRA
low-resource environment
SDG 8 - Decent work and economic growth
status NEW
subject.fl_str_mv Large Language Models (LLM)
Pre-trained models (PTM)
Text classification
News classification
Fine-tuning
Open Source
LoRA
QLoRA
low-resource environment
SDG 8 - Decent work and economic growth
title Open-Source Language Models for News Classification: Implementing Small Models in Low-Resource Environments
title_full Open-Source Language Models for News Classification: Implementing Small Models in Low-Resource Environments
title_fullStr Open-Source Language Models for News Classification: Implementing Small Models in Low-Resource Environments
Open-Source Language Models for News Classification: Implementing Small Models in Low-Resource Environments
title_full_unstemmed Open-Source Language Models for News Classification: Implementing Small Models in Low-Resource Environments
Open-Source Language Models for News Classification: Implementing Small Models in Low-Resource Environments
title_short Open-Source Language Models for News Classification: Implementing Small Models in Low-Resource Environments
title_sort Open-Source Language Models for News Classification: Implementing Small Models in Low-Resource Environments
topic Large Language Models (LLM)
Pre-trained models (PTM)
Text classification
News classification
Fine-tuning
Open Source
LoRA
QLoRA
low-resource environment
SDG 8 - Decent work and economic growth
topic_facet Large Language Models (LLM)
Pre-trained models (PTM)
Text classification
News classification
Fine-tuning
Open Source
LoRA
QLoRA
low-resource environment
SDG 8 - Decent work and economic growth
url http://hdl.handle.net/10362/175829
visible 1