Mafiascum数据集

Mafiascum数据集是由荷兰奈梅亨大学的计算机科学系创建，包含超过700场Mafia游戏的论坛交流记录，总计超过9000份文档。每份文档包含单个玩家在一场游戏中的所有消息。数据集通过随机分配玩家为欺骗性或非欺骗性角色来模拟自然语言中的欺骗行为。该数据集主要用于构建基于先前欺骗研究的精选语言特征集，并测试多种语言欺骗线索的有效性。应用领域包括情报机构、执法和在线市场中的自动欺骗检测。

The Mafiascum Dataset was created by the Department of Computer Science of Radboud University Nijmegen in the Netherlands. It contains forum chat logs from over 700 Mafia games, totaling more than 9,000 documents. Each document includes all messages sent by a single player during one game. The dataset simulates deceptive behavior in natural language by randomly assigning players to either deceptive or non-deceptive roles. It is mainly used to build curated linguistic feature sets based on prior deception research and test the effectiveness of various linguistic deception cues. Its application fields include automated deception detection in intelligence agencies, law enforcement and online marketplaces.