Reasoning under Ambiguity: Uncertainty-Aware Multilingual Emotion Classification under Partial Supervision

TL;DR

The paper tackles multilingual emotion classification under partial supervision by explicitly modeling annotation ambiguity. It introduces the Reasoning under Ambiguity framework, which uses a shared multilingual encoder, entropy-based instability weighting, and a mask-aware objective with positive-unlabeled regularization to robustly learn from incomplete labels. Empirical results on English, Spanish, and Arabic demonstrate improved accuracy, calibration, and interpretability, with ambiguity weighting providing stable gains across languages. The work highlights the importance of aligning learning objectives with the annotation process to achieve trustworthy, multilingual emotion analysis in real-world settings.

Abstract

Contemporary knowledge-based systems increasingly rely on multilingual emotion identification to support intelligent decision-making, yet they face major challenges due to emotional ambiguity and incomplete supervision. Emotion recognition from text is inherently uncertain because multiple emotional states often co-occur and emotion annotations are frequently missing or heterogeneous. Most existing multi-label emotion classification methods assume fully observed labels and rely on deterministic learning objectives, which can lead to biased learning and unreliable predictions under partial supervision. This paper introduces Reasoning under Ambiguity, an uncertainty-aware framework for multilingual multi-label emotion classification that explicitly aligns learning with annotation uncertainty. The proposed approach uses a shared multilingual encoder with language-specific optimization and an entropy-based ambiguity weighting mechanism that down-weights highly ambiguous training instances rather than treating missing labels as negative evidence. A mask-aware objective with positive-unlabeled regularization is further incorporated to enable robust learning under partial supervision. Experiments on English, Spanish, and Arabic emotion classification benchmarks demonstrate consistent improvements over strong baselines across multiple evaluation metrics, along with improved training stability, robustness to annotation sparsity, and enhanced interpretability.

Figures (2)

• Figure 1: Overview of the proposed Reasoning under Ambiguity framework. Multilingual inputs (English, Spanish, Arabic) are encoded using a shared multilingual encoder (XLM-R or mDeBERTa). The [CLS] representation is fed into an ambiguity-aware head that estimates prediction entropy and assigns sample-level weights for ambiguity-weighted learning under partial supervision. An evidential head is included for comparative uncertainty modeling. Missing labels are handled using a mask-aware objective with positive--unlabeled regularization.
• Figure 2: Interpretability under ambiguity (test set). For each instance, we report predicted labels with probabilities, the ambiguity score (entropy $H$), and the derived training influence weight $w=\exp(-\tau H)$. Token-level highlights are computed using Gradient $\times$ Input for each predicted label. Darker red indicates higher token contribution to the corresponding emotion. Across languages, ambiguous instances yield higher entropy and lower weights, while token attributions identify salient lexical, emoji, or contextual cues supporting multi-label predictions.