Scholarlink

<p>In this paper, description of a multimodal empathy detection method using capsule based cross modal entanglement, temporal spatial relational graph modeling, hierarchical attribution, and multi pivot self distillation is discussed for different scenarios. With synchronized textual, aural, and visual signals, the model portrays fine grained emotional dynamics with greater structural coherence sets. Gradient guided attribution maps increase time interpretation, while a graph driven temporal reasoning module captures modality specific transitions. A fused representation latent synchronization metric evaluates emotional trajectory geometric alignment. SEWA and MELD outperform multimodal baselines in accuracy, F1 score, and latent alignment. The concept underpins interpretable empathy detection in complex conversations. Empirical results demonstrate how much the performance improved, from a boost of 4.5% in AUC to over 7.4% in accuracy with low training data while preserving attribution fidelity and latent alignment. Hence, this work establishes an explainable and generalizable foundation for recognition of empathy in a multimodal affective system sets.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Design of an integrated new method for multimodal empathy recognition using capsule entanglement, temporal graphs, and self-distillation frameworks

  • 期刊论文

  • 发表日期：

    2026年1月10日

查看全文
  • Rajesh Aniruddha Tiwari,
  • Sohel A. Bhura

摘要

In this paper, description of a multimodal empathy detection method using capsule based cross modal entanglement, temporal spatial relational graph modeling, hierarchical attribution, and multi pivot self distillation is discussed for different scenarios. With synchronized textual, aural, and visual signals, the model portrays fine grained emotional dynamics with greater structural coherence sets. Gradient guided attribution maps increase time interpretation, while a graph driven temporal reasoning module captures modality specific transitions. A fused representation latent synchronization metric evaluates emotional trajectory geometric alignment. SEWA and MELD outperform multimodal baselines in accuracy, F1 score, and latent alignment. The concept underpins interpretable empathy detection in complex conversations. Empirical results demonstrate how much the performance improved, from a boost of 4.5% in AUC to over 7.4% in accuracy with low training data while preserving attribution fidelity and latent alignment. Hence, this work establishes an explainable and generalizable foundation for recognition of empathy in a multimodal affective system sets.