Transforming sports knowledge: Chatbot AI-based systems for improved training and education in sports
Journal of Physical Education and Sport · 31.01.2026 · 26 (1) · 130-139
DOI: 10.7752/jpes.2026.01015 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Physical Education and Sport, online erschienen am 31.01.2026. Verfasst von Tri Lathif Mardi Suryanto und 5 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 3.0).
Abstract (Originalsprache)
This study introduces a Multiple Closed-Domain (MCD) Question Answering (QA) system utilizing Transformer-based models to optimize access to sports knowledge, improving training and management in the sports sector. Traditional QA systems, whether closed-domain or open-domain, each come with inherent limitations: closed-domain systems excel in accuracy within specific topics but lack scalability, while opendomain systems offer broader coverage but often generate irrelevant or incorrect responses. To overcome these challenges, this research proposes an optimized MCD QA framework using Transformer models, specifically BART-base and BART-large, designed to address the needs of sports communities by providing accurate, context-aware answers across multiple closed domains. A dataset of 10,000 question-answer pairs was curated from five major Indonesian sports federations—PSSI (football), PBESI (e-sports), PBVSI (volleyball), PBSI (badminton), and PERBASI (basketball). The data was cleaned, tokenized, and validated by experts to ensure its relevance to sports-specific queries. Hyperparameter optimization, performed using Optuna, significantly improved the performance of the models. Evaluation metrics, including ROUGE, BLEU, and METEOR, assessed the quality of the generated responses in terms of fluency, accuracy, and semantic relevance. The results show that the optimized models achieved superior performance, particularly the BART-large model, which demonstrated the highest ROUGE-L score, indicating its ability to capture essential sports knowledge. The BART-base model excelled in BLEU and METEOR, highlighting its fluency and semantic coherence. This system provides a scalable, efficient solution for accessing sports regulations, training techniques, and health policies, essential for athletes, coaches, and sports managers. AI-driven system has significant potential in sports education and management, enabling quicker access to accurate information, improving decision-making processes, and enhancing the overall quality of sports training and performance.
Zitieren
APA
Suryanto, T. L. M., Wahyuni, E. D., Sa’diyah, I., Wibawa, A. P., Sulistyo, D. A., & Haqiyah, A. (2026). Transforming sports knowledge: Chatbot AI-based systems for improved training and education in sports. Journal of Physical Education and Sport, 26(1), 130-139. https://doi.org/10.7752/jpes.2026.01015
Vancouver
Suryanto TLM, Wahyuni ED, Sa’diyah I, Wibawa AP, Sulistyo DA, Haqiyah A. Transforming sports knowledge: Chatbot AI-based systems for improved training and education in sports. Journal of Physical Education and Sport. 2026;26(1):130-139. doi:10.7752/jpes.2026.01015
BibTeX
@article{Suryanto2026,
author = {Suryanto, Tri Lathif Mardi and Wahyuni, Eka Dyar and Sa’diyah, Ilmatus and Wibawa, Aji Prasetya and Sulistyo, Danang Arbian and Haqiyah, Aridhotul},
title = {Transforming sports knowledge: Chatbot AI-based systems for improved training and education in sports},
journal = {Journal of Physical Education and Sport},
year = {2026},
volume = {26},
number = {1},
pages = {130--139},
doi = {10.7752/jpes.2026.01015}
}
EUVIMED