Mafiascum

超过 700 款 Mafia 游戏的集合，其中玩家被随机分配欺骗性或非欺骗性角色，然后通过论坛帖子进行互动。从数据集中编译了 9000 多个文档，每个文档都包含单个玩家在单个游戏中编写的所有消息。该语料库用于构建一组基于先前欺骗研究的精心挑选的语言特征，以及一组富含子词信息的平均词向量。A collection of over 700 Mafia games, where players are randomly assigned either deceptive or non-deceptive roles and interact via forum posts. More than 9,000 documents were compiled from this dataset, with each document containing all messages written by a single player in a single game. This corpus is used to construct a carefully selected set of linguistic features based on prior deception research, as well as a set of average word embeddings rich in subword information.

搜集汇总

背景与挑战

背景概述

Mafiascum数据集是一个包含700多款Mafia游戏的大型文本语料库，其中玩家被随机分配欺骗性或非欺骗性角色，并通过论坛帖子互动。该数据集编译了9000多个文档，每个文档代表单个玩家在游戏中的所有消息，用于构建语言特征和词向量以支持欺骗检测研究。