Ifemma/pidginshield

该数据集名为pidginshield，旨在检测尼日利亚皮钦语、英语和混合语言中的加密货币诈骗。它捕捉了西非加密货币社区中常见的诈骗模式，这些社区广泛使用非正式语言和俚语。数据集包含文本、语言类型、标签（诈骗或合法）、诈骗类别、意图和解释等字段。使用场景包括训练诈骗检测模型、提高金融科技和Web3中的AI安全性，以及为加密货币社区提供审核工具。数据集由Adaption Labs为Uncharted Data Challenge创建，并利用了Adaption的数据基础设施。

this dataset is designed to detect crypto scams in nigerian pidgin, english, and mixed-language communication. it captures realistic scam patterns commonly found in west african crypto communities, where informal language and slang are widely used. each entry contains: text, language_type, label (scam or legit), scam_category, intent, and explanation. use cases include training scam detection models, improving ai safety in fintech and web3, and moderation tools for crypto communities. the dataset, pidginshield, was created for the uncharted data challenge by adaption labs and leverages adaptive data, adaption’s data infrastructure for building and evolving datasets.