Temporal fine-tuning for early risk detection

Early Risk Detection (ERD) on the Web aims to identify promptly users facing social and health issues. Users are analyzed post-by-post, and it is necessary to guarantee correct and quick answers, which is particularly challenging in critical scenarios. ERD involves optimizing classification precisio...

Descripción completa

Guardado en:
Detalles Bibliográficos
Autores principales: Thompson, Horacio, Villatoro-Tello, Esaú, Montes-y-Gómez, Manuel, Errecalde, Marcelo Luis
Formato: Objeto de conferencia
Lenguaje:Inglés
Publicado: 2024
Materias:
Ciencias Informáticas
Intelligent Systems
Machine Learning
Transformers
Early Risk Detection
Mental Health
Acceso en línea:http://sedici.unlp.edu.ar/handle/10915/178838
Aporte de:
SEDICI (UNLP) de Universidad Nacional de La Plata
id I19-R120-10915-178838
record_format dspace
spelling I19-R120-10915-1788382025-05-08T20:07:47Z http://sedici.unlp.edu.ar/handle/10915/178838 Temporal fine-tuning for early risk detection Thompson, Horacio Villatoro-Tello, Esaú Montes-y-Gómez, Manuel Errecalde, Marcelo Luis 2024-08 2024 2025-05-08T17:04:15Z en Ciencias Informáticas Intelligent Systems Machine Learning Transformers Early Risk Detection Mental Health Early Risk Detection (ERD) on the Web aims to identify promptly users facing social and health issues. Users are analyzed post-by-post, and it is necessary to guarantee correct and quick answers, which is particularly challenging in critical scenarios. ERD involves optimizing classification precision and minimizing detection delay. Standard classification metrics may not suffice, resorting to specific metrics such as ERDEθ that explicitly consider precision and delay. The current research focuses on applying a multi-objective approach, prioritizing classification performance and establishing a separate criterion for decision time. In this work, we propose a completely different strategy, temporalfine-tuning, which allows tuning transformer-based models by explicitly incorporating time within the learning process. Our method allows us to analyze complete user post histories, tune models considering different contexts, and evaluate training performance using temporal metrics. We evaluated our proposal in the depression and eating disorders tasks for the Spanish language, achieving competitive results compared to the best models of MentalRiskES 2023. We found that temporal fine-tuning optimized decisions considering context and time progress. In this way, by properly taking advantage of the power of transformers, it is possible to address ERD by combining precision and speed as a single objective. Sociedad Argentina de Informática e Investigación Operativa Objeto de conferencia Objeto de conferencia http://creativecommons.org/licenses/by-nc-sa/4.0/ Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) application/pdf 137-149
institution Universidad Nacional de La Plata
institution_str I-19
repository_str R-120
collection SEDICI (UNLP)
language Inglés
topic Ciencias Informáticas
Intelligent Systems
Machine Learning
Transformers
Early Risk Detection
Mental Health
spellingShingle Ciencias Informáticas
Intelligent Systems
Machine Learning
Transformers
Early Risk Detection
Mental Health
Thompson, Horacio
Villatoro-Tello, Esaú
Montes-y-Gómez, Manuel
Errecalde, Marcelo Luis
Temporal fine-tuning for early risk detection
topic_facet Ciencias Informáticas
Intelligent Systems
Machine Learning
Transformers
Early Risk Detection
Mental Health
description Early Risk Detection (ERD) on the Web aims to identify promptly users facing social and health issues. Users are analyzed post-by-post, and it is necessary to guarantee correct and quick answers, which is particularly challenging in critical scenarios. ERD involves optimizing classification precision and minimizing detection delay. Standard classification metrics may not suffice, resorting to specific metrics such as ERDEθ that explicitly consider precision and delay. The current research focuses on applying a multi-objective approach, prioritizing classification performance and establishing a separate criterion for decision time. In this work, we propose a completely different strategy, temporalfine-tuning, which allows tuning transformer-based models by explicitly incorporating time within the learning process. Our method allows us to analyze complete user post histories, tune models considering different contexts, and evaluate training performance using temporal metrics. We evaluated our proposal in the depression and eating disorders tasks for the Spanish language, achieving competitive results compared to the best models of MentalRiskES 2023. We found that temporal fine-tuning optimized decisions considering context and time progress. In this way, by properly taking advantage of the power of transformers, it is possible to address ERD by combining precision and speed as a single objective.
format Objeto de conferencia
Objeto de conferencia
author Thompson, Horacio
Villatoro-Tello, Esaú
Montes-y-Gómez, Manuel
Errecalde, Marcelo Luis
author_facet Thompson, Horacio
Villatoro-Tello, Esaú
Montes-y-Gómez, Manuel
Errecalde, Marcelo Luis
author_sort Thompson, Horacio
title Temporal fine-tuning for early risk detection
title_short Temporal fine-tuning for early risk detection
title_full Temporal fine-tuning for early risk detection
title_fullStr Temporal fine-tuning for early risk detection
title_full_unstemmed Temporal fine-tuning for early risk detection
title_sort temporal fine-tuning for early risk detection
publishDate 2024
url http://sedici.unlp.edu.ar/handle/10915/178838
work_keys_str_mv AT thompsonhoracio temporalfinetuningforearlyriskdetection
AT villatorotelloesau temporalfinetuningforearlyriskdetection
AT montesygomezmanuel temporalfinetuningforearlyriskdetection
AT errecaldemarceloluis temporalfinetuningforearlyriskdetection
_version_ 1847925379136749568

Ejemplares similares