Dialogue Safety in Mental Health Support Benchmark
收藏资源简介:
本数据集名为‘心理健康支持对话安全性基准’,由浙江大学和西湖大学工学院联合创建。数据集包含7935条多轮对话,来源于一个中文在线文本咨询平台,每条对话均由寻求帮助者和经验丰富的支持者进行。数据集的创建旨在通过细粒度标签分析对话安全性,特别是在心理健康支持领域。创建过程中,使用了包括BERT-base、RoBERTa-large和ChatGPT在内的流行语言模型进行分析。该数据集的应用领域主要集中在改善对话系统在心理健康支持中的安全性和有效性,确保对话对寻求帮助者产生积极影响。
This dataset, titled "Safety Benchmark for Mental Health Support Conversations", was jointly developed by Zhejiang University and the School of Engineering of Westlake University. It contains 7,935 multi-turn conversations sourced from a Chinese online text-based counseling platform, with each conversation involving a help-seeker and an experienced support professional. The dataset was created to analyze the safety of conversations in the mental health support domain using fine-grained labeling. During the dataset construction process, prevalent language models including BERT-base, RoBERTa-large, and ChatGPT were utilized for analysis. The primary application scenarios of this dataset focus on enhancing the safety and effectiveness of dialogue systems in mental health support, with the goal of ensuring that conversations deliver positive impacts to help-seeking individuals.

- 1A Benchmark for Understanding Dialogue Safety in Mental Health Support浙江大学 西湖大学工学院 · 2023年



