Situational Estimation of Sports Broadcasting using a Character Level Auto-Encoder for Live Tweets
Nodoka Fujimoto, Taketoshi Ushiama · 2020
Real-time videos enable people to experience real-world events in real time. However, it is challenging for people to keep watching an event when they are busy and have other activities to carry out. It is therefore crucial to develop a method for supporting the viewers' real-time video viewing effectively. Using sports broadcasting as a representative example of such videos, our goal is to develop a system that detects and automatically estimates the events in sports games that are likely to be of high value to the viewers and notifies them in real time during sports broadcasts. To this end, we propose a general-purpose method that efficiently models situations from small amounts of tweets without using domain-specific knowledge. The effectiveness of the method was verified via experiments by estimating the number of tweets posted from the modeled features of a situation with high accuracy.