DianJin/DianJin-CSC-Data

CSConv是一个基于真实世界客户与客服人员对话的评价数据集，通过大型语言模型重写，以体现明确的策略使用，并进行了相应的标注。RoleCS是一个角色扮演的训练数据集，使用大型语言模型生成，与CSC框架对齐的富有策略性的对话。这两个数据集旨在支持客户支持对话系统的模型开发、基准测试和进一步研究。

CSConv is an evaluation dataset of real-world customer-agent conversations rewritten using large language models to reflect deliberate strategy use, and it is annotated accordingly. RoleCS is a role-playing training dataset that uses large language models to generate strategy-rich conversations aligned with the CSC framework. These two datasets are intended to support the development of customer support conversation systems, benchmarking, and further research in the field.