PsyDTCorpus 心理咨询师数字孪生数据集

PsyDTCorpus 是一个由华南理工大学未来技术学院-广东省数字孪生人重点实验室于 2024 年推出的心理咨询师数字孪生数据集。该数据集的核心目标是模拟特定心理咨询师的语言风格和咨询技术，以支持心理咨询师数字孪生大模型 SoulChat2.0 的开发和训练。相关论文成果为「SoulChat: Improving LLMs’ Empathy, Listening, and Comfort Abilities through Fine-tuning with Multi-turn Empathy Conversations」。

PsyDTCorpus is a digital twin dataset for psychological counselors launched in 2024 by the Guangdong Provincial Key Laboratory of Digital Twin Humans, School of Future Technology, South China University of Technology. The core objective of this dataset is to simulate the linguistic style and counseling techniques of specific psychological counselors, so as to support the development and training of SoulChat2.0, a large language model for psychological counselor digital twins. The associated research paper is titled "SoulChat: Improving LLMs’ Empathy, Listening, and Comfort Abilities through Fine-tuning with Multi-turn Empathy Conversations".

搜集汇总

背景与挑战

背景概述

PsyDTCorpus是华南理工大学2024年推出的心理咨询师数字孪生数据集，旨在模拟特定咨询师的语言风格和疗法技术，支持SoulChat2.0模型的训练。该数据集基于5k单轮咨询样本合成了5k高质量心理健康对话数据，总轮次达90,365轮，并在谈话技术等四个专业维度上通过人工评估验证了其有效性。