Building Emotional Support Chatbots in the Era of LLMs
Zhonghua Zheng1 Lizi Liao2 Yang Deng3 Liqiang Nie1 1 Harbin Institute of Technology, Shenzhen 2 Singapore Management University 3 National University of Singapore {polang1999,
Abstract
While emotional support in conversational scenarios offers societal benefits, limited data and non-standardized training impede its application. This work endeavors to navigate these challenges by harnessing the capabilities of Large Language Models (LLMs). We introduce an innovative methodology that synthesizes human insights with the computational prowess of LLMs to curate an extensive emotional support dialogue dataset. Our approach is initiated with a meticulously designed set of dialogues spanning diverse scenarios as generative seeds. By utilizing the in-context learning potential of ChatGPT, we recursively generate an ExTensible Emotional Support dialogue dataset, named ExTES. Following this, we deploy advanced tuning techniques on the LLaMA model, examining the impact of diverse training strategies, ultimately yielding an LLM meticulously optimized for emotional support interactions. An exhaustive assessment of the resultant model showcases its proficiency in offering emotional support, marking a pivotal step in the realm of emotional support bots and paving the way for subsequent research and implementations. The dataset and codes are available here111https://anonymous.4op
中文速览
用大规模语言模型（LLM）来构建情感支持对话系统，一直面临两大瓶颈：高质量标注数据太少，以及缺乏规范的训练方法。研究团队提出了一套"人机协作"的数据生成流程：先由人工精心设计涵盖36种生活场景和16种支持策略的种子对话，再利用ChatGPT的上下文学习能力反复自我对话、迭代扩充，最终构建出名为ExTES的大规模情感支持对话数据集，且只有不到10%的生成对话需要人工修正，大幅降低了人力成本。在此基础上，研究者对开源模型LLaMA采用参数高效微调等多种训练策略进行优化，并通过自动评估和人工评估全面验证效果。结果表明，ExTES在多个维度上显著优于现有的ESConv数据集，基于ExTES微调的LLaMA也展现出出色的情感支持能力——这项工作为心理健康辅导、社交互动等场景中的情感支持机器人提供了可扩展的数据基础和可复现的训练范式。
原文 arXiv:2308.11584;中英对照 + 大白话阅读 https://aha.fim.ai/paper/2308.11584v1