NTDH:面向综合情感分析的复杂推理方法
原标题:NTDH: Complex Reasoning for Comprehensive Affective Analysis
AI 摘要
arXiv 论文提出 NTDH 方法,将综合情感分析重构为复杂推理问题,通过自然化、容差感知门控、领域感知策略和方向性提示解决推理轨迹合成中的对齐与失败案例问题。使用 Qwen3-8B 模型,结合 SFT 和 GRPO 训练,仅用 16,302 条训练记录(比同类系统少约 14 倍),在六个官方测试指标中的五个上优于 SFT 基线,并在 EI-reg 任务上取得最强结果(Pearson 相关系数 0.862)。
正文节选
Computer Science > Computation and Language Title:NTDH: Complex Reasoning for Comprehensive Affective Analysis View PDF HTML (experimental) Abstract:Comprehensive affective analysis is challenging for two reasons: it spans heterogeneous prediction tasks with continuous, ordinal, and multi-label outputs, and affective meaning is context-dependent, requiring conflicting cues to be reconciled rather than mapped directly to labels. Existing methods learn this mapping directly and do not