AUTOMATED ANALYSIS OF CROSSING ACTIONS IN FOOTBALL COMMENTARY USING LARGE LANGUAGE MODELS

Date

2024-4-19

Author

Erkul, Anıl

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

53
views

14
downloads

Cite This

In football, well-executed crosses have a significant impact, directly creating about 15% of scoring opportunities in the top leagues of England, Italy, and France. This thesis focuses on the development of a comprehensive dataset sourced from the in-depth analysis of football commentary extracted from the English Premier League matches spanning the 2022-2023 Football season. The main goal is to gather and organize key details related to crossing actions. This involves collecting diverse information, such as the outcome of each cross, identifying the team and the player making the cross, and evaluating the sentiment or qualitative aspects of the executed crosses with human annotators. In the second phase of the thesis, large language models are used and fine-tuned to automatically identify, extract, and describe instances of crossing actions, along with their related details, in the extensive football commentary from Premier League matches. The results indicate that the large language models are useful to reveal crossing actions successfully.

Subject Keywords

Football, Crossing, Labeling, Large Language Models, Text Classification

URI

https://hdl.handle.net/11511/109440

Collections

Graduate School of Informatics, Thesis