Learning How Spectator Reactions Affect Popularity on Twitch
Learning How Spectator Reactions Affect Popularity on Twitch
复制标题
了解观众反应如何影响 Twitch 上的受欢迎程度
DOI：
10.1109/bigcomp48618.2020.00-84
复制
发表时间：
2020
期刊：
影响因子：
--
通讯作者：
M. Cha
中科院分区：
文献类型：
--
作者：
Jeongmin Kim;Kunwoo Park;Hyeonho Song;J. Y. Park;M. Cha
游戏直播已成为 Twitch.tv 等服务上需求旺盛的在线娱乐流量。实时聊天与直播流允许观众互相讨论内容。这种新形式的在线内容承载了大量观众的反应，包括文本和表情符号，并带来了预测哪个视频最终流行的独特机会。这项研究提出了一个基于新编译的数据集的观众计数预测模型，该数据集描述了 Twitch 直播的视频流行度和观众的聊天反应。我们的分析表明，从直播早期阶段捕获的观众反应（例如几个小时内容中的前 15 分钟）是视频流最终流行的重要标志。我们讨论研究结果的影响和未来的方向。
Gameplay live streaming has become a highly demanded online entertainment traffic on services like Twitch.tv. The live chat concurrent with on-air streams allows viewers to talk about the content with each other. This new form of online content carries numerous spectator reactions, including text and emojis, and brings a unique opportunity to predict which video becomes popular in the end. This study presents a prediction model of viewer counts on a newly compiled dataset describing the video popularity of Twitch live streams and chat reactions of spectators. Our analysis demonstrates that the spectator reactions captured from the early-stage of live streams, such as the first 15 minutes out of several hours of content, hold essential markers of the eventual popularity of video streams. We discuss the implications of the findings and future directions.