A Multimodal Semantic Model for Event Identification on Sports Media Content
A Multimodal Semantic Model for Event Identification on Sports Media Content

Summary/Abstract: This paper presents a multimodal semantic event identification model of occurring events in sports games. Sport broadcasting programs and TV shows form a demanding media case, as journalists, experts and analysts tend to examine in detail their gameplay aspects and elements, trying to offer thorough analysis and deep insights to their large audience. Basketball serves as an exemplary case for developing an event identification model in sports media content, since it is full of small or bigger activities that take place throughout the game. In this work, a basketball event is defined as a series of player actions during a team possession in the game. The proposed method leverages and combines the information of various game sources as play-by-play and text reports. Given the current state of research, the current paper proposes multilevel event classification taxonomies that could be used in multimodal semantic processing and conceptualization approaches.

• Details
• Contents