Classification of texts describing baseball batting results on Twitter using BERT

• Published in: Studies in Science and Technology Vol. 12; no. 1; pp. 93 - 99
• Main Authors: Ikeda, Riku, Ueda, Yoshihiro, Sakamoto, Kazuma
• Format: Journal Article
• Language: English; Japanese
• Published: Osaka Society for Science and Technology 01.01.2023; Japan Science and Technology Agency
• Subjects: News; Real time; Smartphones; Sports
• ISSN: 2186-4942; 2187-1590
• DOI: 10.11425/sst.12.93

With the improvement of communication speeds and the spread of smartphones and IoT devices, there is an abundance of sports video content on the Internet, and users can watch game broadcasts and highlight videos without any time or location restrictions. Although sports games are very long and difficult to watch continuously, users find it more attractive and valuable to watch them in real-time than to record them. In this research, we develop a breaking news system that automatically identifies a specific player’s cue, results, and chance scenes by understanding the context of real-time tweets on Twitter and sports commentary from commentators using BERT. This system will enable users who have difficulty in watching in real-time or who are only interested in specific players to watch, as well as to automate breaking news that is currently done manually.