NTDH: Complex Reasoning for Comprehensive Affective Analysis

NTDH：用于综合情感分析的复杂推理

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 model the reconciliation explicitly.

综合情感分析面临两大挑战：首先，它涵盖了具有连续、序数和多标签输出的异构预测任务；其次，情感意义具有上下文依赖性，需要对冲突的线索进行调和，而非直接映射到标签。现有的方法大多直接学习这种映射，而没有显式地对调和过程进行建模。

We recast the task as a complex-reasoning problem, which yields one output interface across heterogeneous label spaces and a trajectory over which a verifiable reward can be optimised; to our knowledge, this is the first such treatment covering both sentiment and emotion. The obstacle is on the data side: affective reasoning traces must be synthesised, and generic synthesis is misaligned with the targets, tolerances, and phenomena of affect, and discards or leaks its failure cases.

我们将该任务重构为一个复杂推理问题，从而在异构标签空间中产生统一的输出接口，并形成一条可优化可验证奖励的轨迹；据我们所知，这是首个同时涵盖情感（Sentiment）与情绪（Emotion）的处理方案。目前的障碍在于数据端：情感推理轨迹必须进行合成，而通用的合成方法与情感的目标、容差和现象不匹配，且往往会丢弃或泄露其失败案例。

We propose NTDH, which addresses these four failures. Naturalisation sets the training answer to the gold label, so it is correct by construction. A Tolerance-aware gate checks each answer against the task’s own scoring margin. Domain-aware strategies refine the reasoning using ideas from affective science. Directional Hints report only the type and direction of an error, without exposing the target.

我们提出了 NTDH，旨在解决上述四个缺陷。“自然化”（Naturalisation）将训练答案设定为金标准标签，从而在构建上保证其正确性。“容差感知门”（Tolerance-aware gate）根据任务自身的评分余量检查每个答案。“领域感知策略”（Domain-aware strategies）利用情感科学的理念优化推理过程。“方向性提示”（Directional Hints）仅报告错误的类型和方向，而不泄露目标答案。

We train Qwen3-8B with SFT and then GRPO under the same tolerance used for verification (up to a more permissive construction gate on the multi-label subtask), and a component ablation quantifies the data-quality effect of each part. Using 16,302 training records, about 14x fewer than comparable instruction-tuned systems, the final policy improves over its SFT checkpoint on five of six official-test metrics and achieves the strongest EI-reg result among the compared systems, at a Pearson correlation of 0.862.

我们使用 SFT 对 Qwen3-8B 进行训练，随后在验证时采用相同的容差进行 GRPO 训练（在多标签子任务上采用了更宽松的构建门），并通过组件消融实验量化了各部分对数据质量的影响。仅使用 16,302 条训练记录（比同类指令微调系统少约 14 倍），最终策略在六项官方测试指标中的五项上均优于其 SFT 检查点，并取得了对比系统中最佳的 EI-reg 结果，皮尔逊相关系数达到 0.862。