A Novel Approach to Automated Live-Ticker Generation in Football: Using Large Language Models and Audio Data

James Anurathan, Manfred Rössle, Marco Klaiber

2025

Abstract

Football (soccer) is one of the most popular sports in the world, with fans enjoying real-time coverage of their favorite team’s from anywhere. Explicitly, the progress in the field of Artificial Intelligence (AI) holds great potential to further improve this experience and optimize the delivery of content. In this context, our work investigates the integration of Large Language Models (LLMs) – in our case GPT-4 – with Advanced Speech Recognition (ASR) systems to automate the creation of live football ticker commentary. For this purpose, we present an approach for transcribing live audio commentary from real football matches, utilizing a whisper model to prepare the transcribed text for correct input to the LLM. This approach is leveraged by Named Entity Recognition (NER) and BERT-based models to provide clear, user-friendly, and multilingual texts for live tickers. In addition, we evaluate our approach with an objective and metric-based method to transparently assess the effectiveness of our approach. The study shows the potential of LLMs in automating sports commentary, but also emphasizes the importance of refining entity recognition and addressing content accuracy issues. Future work should focus on improving transcription accuracy, refining NER models, and mitigating LLM hallucinations to develop more reliable and scalable automated live ticker systems.

Paper Citation

in Harvard Style

Anurathan J., Rössle M. and Klaiber M. (2025). A Novel Approach to Automated Live-Ticker Generation in Football: Using Large Language Models and Audio Data. In Proceedings of the 13th International Conference on Sport Sciences Research and Technology Support - Volume 1: icSPORTS; ISBN 978-989-758-771-9, SciTePress, pages 134-141. DOI: 10.5220/0013665700003988

in Bibtex Style

@conference{icsports25,
author={James Anurathan and Manfred Rössle and Marco Klaiber},
title={A Novel Approach to Automated Live-Ticker Generation in Football: Using Large Language Models and Audio Data},
booktitle={Proceedings of the 13th International Conference on Sport Sciences Research and Technology Support - Volume 1: icSPORTS},
year={2025},
pages={134-141},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0013665700003988},
isbn={978-989-758-771-9},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 13th International Conference on Sport Sciences Research and Technology Support - Volume 1: icSPORTS
TI - A Novel Approach to Automated Live-Ticker Generation in Football: Using Large Language Models and Audio Data
SN - 978-989-758-771-9
AU - Anurathan J.
AU - Rössle M.
AU - Klaiber M.
PY - 2025
SP - 134
EP - 141
DO - 10.5220/0013665700003988
PB - SciTePress