A Feature-Oriented Textual Ensemble Framework for Robust Emotion Detection Across Diverse Classification Paradigms
International Journal of Computational and Experimental Science and Engineering · 30.04.2025 · 11 (4)
DOI: 10.22399/ijcesen.4934 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Computational and Experimental Science and Engineering, online erschienen am 30.04.2025. Verfasst von Mukesh C Jain und 1 weiteren Autor:innen. Die Arbeit nennt 47 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Emotion detection from text plays a crucial role in natural language processing applications such as social media analysis, opinion mining, and conversational systems. Existing approaches typically rely on isolated representation paradigms, including lexical features, semantic embeddings, or contextualized models, which often limits robustness across datasets and task formulations. This study presents a Feature-Oriented Textual Ensemble (FOTE) framework that integrates lexical, semantic, and contextual representations within a unified probability-level ensemble architecture. Unlike conventional ensembles that emphasize model-level diversity, the proposed framework explicitly combines heterogeneous representation paradigms designed to capture complementary linguistic properties. Parallel representation branches are employed to model explicit emotion cues, sentence-level semantics, and context-dependent word meaning, followed by probability-level fusion using ensemble strategies, while maintaining a consistent implementation pipeline across experimental settings. Configuration-level adaptations are applied to support binary, multi-class, and multi-label emotion classification without architectural modification. The proposed framework is evaluated on multiple benchmark and validation datasets spanning diverse domains and emotion classification paradigms. Experimental results indicate that the ensemble consistently improves performance over a strong transformer-based baseline across most datasets, demonstrating enhanced robustness and generalization, while avoiding performance degradation in challenging scenarios. The results highlight the effectiveness of representation-level integration as a generalizable alternative to single-paradigm emotion detection models.
Zitieren
APA
Jain, M. C., & Haneef, F. (2025). A Feature-Oriented Textual Ensemble Framework for Robust Emotion Detection Across Diverse Classification Paradigms. International Journal of Computational and Experimental Science and Engineering, 11(4). https://doi.org/10.22399/ijcesen.4934
Vancouver
Jain MC, Haneef F. A Feature-Oriented Textual Ensemble Framework for Robust Emotion Detection Across Diverse Classification Paradigms. International Journal of Computational and Experimental Science and Engineering. 2025;11(4). doi:10.22399/ijcesen.4934
BibTeX
@article{Jain2025,
author = {Jain, Mukesh C and Haneef, Farha},
title = {A Feature-Oriented Textual Ensemble Framework for Robust Emotion Detection Across Diverse Classification Paradigms},
journal = {International Journal of Computational and Experimental Science and Engineering},
year = {2025},
volume = {11},
number = {4},
doi = {10.22399/ijcesen.4934}
}
EUVIMED