GameFlow: Narrative Visualization of NBA Basketball Games.
Wei Chen,Tianyi Lao,Jing Xia,Xinxin Huang,Biao Zhu,Wanqi Hu,Huihua Guan
DOI: https://doi.org/10.1109/TMM.2016.2614221
2016-01-01
Abstract:Although basketball games have received broad attention, the forms of game reports and webcast are purely content-based cross-media: texts, videos, snapshots, and performance figures. Analytical narrations of games that seek to compose a complete game from heterogeneous datasets are challenging for general media producers because such a composition is time-consuming and heavily depends on domain experts. In particular, an appropriate analytical commentary of basketball games requires two factors, namely, rich context and domain knowledge, which includes game events, player locations, player profiles, and team profiles, among others. This type of analytical commentary elicits a timely and effective basketball game data visualization made up of different sources of media. Existing visualizations of basketball games mainly profile a particular aspect of the game. Therefore, this paper presents an expressive visualization scheme that comprehensively illustrates NBA games with three levels of details: a season level, a game level, and a session level. We reorganize a basketball game as a sequence of sessions to depict the game states and heated confrontations. We design and implement a live system that integrates multimedia NBA datasets: play-by-play text data, box score data, game video data, and action area data. We demonstrate the effectiveness of this scheme with case studies and user feedbacks.