情境-情绪感知的治疗对话生成：一种多组件强化学习方法用于语言模型的心理健康支持
Context-Emotion Aware Therapeutic Dialogue Generation: A Multi-component Reinforcement Learning Approach to Language Models for Mental Health Support
AI总结：
本文提出一种多组件强化学习方法，通过改进GPT-2的治疗对话生成能力，提升情绪准确性和上下文相关性，以支持心理健康领域的发展。
摘要：
心理健康障碍对全球社会经济造成了重大负担。尽管大型语言模型（LLMs）提供了24/7、非评判性的互动来填补这一空白，但预训练模型缺乏情境连贯性和情绪一致性，无法产生适当的治疗对话。现有方法存在三个关键方法论缺口：1）监督微调（SFT）会产生重复、缺乏上下文敏感性的输出，无法在临床准确性与真实同理心之间取得平衡；2）基于强化学习（RL）的治疗系统依赖于通用奖励函数（如BLEU、ROUGE），这些函数优先考虑词汇相似性而非临床特定的情绪适当性和上下文相关性；3）LLMs资源消耗大且存在数据隐私风险，使在临床环境中本地部署不可行。为了解决这些缺口，本研究探讨了将SFT和RL技术应用于增强GPT-2的治疗对话生成能力。方法重新组织了输入格式，以同时处理上下文信息和情绪状态以及用户输入，采用了一种新的多组件奖励函数，该函数明确将模型输出与专业治疗逻辑（而不仅仅是词汇重叠）和标注情绪对齐。结果显示，通过RL在多个评估指标上均显著优于基线GPT-2：BLEU（0.0111）、ROUGE-1（0.1397）、ROUGE-2（0.0213）、ROUGE-L（0.1317）和METEOR（0.0581）。LLM评估证实了高上下文相关性和专业性，而RL在情绪准确性上达到了99.34%，相比基线GPT-2的66.96%。这些发现表明RL在开发能够为治疗师提供有价值辅助工具的治疗对话系统方面是有效的，同时保持必要的临床监督。
英文摘要：
Mental health disorders impose a substantial global socioeconomic burden. While large language models (LLMs) offer 24/7, non-judgmental interactions to address this gap, pretrained models lack contextual coherence and emotional alignment for appropriate therapeutic dialogue. Existing methods suffer from three critical methodological gaps: 1) Supervised Fine-Tuning (SFT) produces repetitive, context-insensitive outputs that fail to balance clinical accuracy with genuine empathy; 2) Reinforcement Learning (RL)-based therapeutic systems rely on generic reward functions (e.g., BLEU, ROUGE) that prioritise lexical similarity over clinical-specific emotional appropriateness and contextual relevance; 3) LLMs are resource-intensive and pose data privacy risks, making local deployment in clinical settings infeasible. To address these gaps, this study investigates the application of SFT and RL techniques to enhance GPT-2's capacity for therapeutic dialogue generation. The methodology restructured input formats to enable simultaneous processing of contextual information and emotional states alongside user input, employing a novel multi-component reward function that explicitly aligns model outputs with professional therapeutic logic (not just lexical overlap) and annotated emotions. Results demonstrated substantial improvements through RLs over baseline GPT-2 across multiple evaluation metrics: BLEU (0.0111), ROUGE-1 (0.1397), ROUGE-2 (0.0213), ROUGE-L (0.1317), and METEOR (0.0581). LLM evaluation confirmed high contextual relevance and professionalism, while RL achieved 99.34% emotion accuracy compared to 66.96% for baseline GPT-2. These findings demonstrate RL's effectiveness in developing therapeutic dialogue systems that can serve as valuable assistive tools for therapists, while maintaining essential human clinical oversight.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
