just-ai/jayguard-ner-benchmark

Jay Guard NER Benchmark是一个俄语数据集，用于评估命名实体识别模型在识别个人和敏感数据方面的性能。该数据集来源于现实世界的复杂对话文本，包括工作聊天、客户支持日志和口头语言记录。数据集特别关注保护个人数据的实体，如人名和街道地址等。数据集分为训练集、验证集和测试集，每个实例包括令牌列表和对应的命名实体标签列表。

The Jay Guard NER Benchmark is a Russian-language dataset designed for evaluating Named Entity Recognition (NER) models on their ability to identify personal and sensitive data. The dataset is sourced from real-world, complex conversational texts, including work chats, customer support logs, and spoken language transcripts. It specifically focuses on entities critical for personal data protection, such as `PERSON` and `STREET_ADDRESS`. The dataset is split into train, validation, and test sets, with each instance consisting of a list of `tokens` and a corresponding list of `ner_tags`.