Game-MUG
收藏资源简介:
Game-MUG是由悉尼大学开发的多模态游戏情境理解和评论生成数据集,专注于电子竞技领域。该数据集从2020至2022年的《英雄联盟》游戏直播中收集,包含文本、音频和时间序列事件日志等多种数据类型,旨在帮助理解游戏情境并生成观众参与的评论。数据集包含70,711个转录句子和3,657,611个聊天实例,覆盖了从腾讯英雄联盟职业联赛到世界锦标赛等多个顶级联赛的比赛。通过整合观众讨论、情绪和特定领域信息,Game-MUG支持开发能够全面理解游戏情境并生成类似人类评论的模型,从而增强观众的参与感和理解度。
Game-MUG is a multimodal game scenario understanding and comment generation dataset developed by the University of Sydney, focusing on the esports domain. This dataset is collected from *League of Legends* live broadcasts spanning 2020 to 2022, including multiple data types such as text, audio, and time-series event logs. It aims to facilitate the understanding of game scenarios and the generation of audience-engaging comments. The dataset contains 70,711 transcribed sentences and 3,657,611 chat instances, covering multiple top-tier competitions ranging from the Tencent League of Legends Pro League to the League of Legends World Championship. By integrating audience discussions, emotional information, and domain-specific knowledge, Game-MUG supports the development of models that can comprehensively understand game scenarios and generate human-like comments, thereby enhancing audience engagement and comprehension.




