EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Beyond Polarity: A Scalable Transformer-Based Emotion Intelligence Framework With Distributed Big Data Optimization

K. Adi Narayana Reddy, Shadab Ahmad, Pandi Chiranjeevi, Thatikonda Supraja

IEEE Access · 2026 · 14 · 101726-101745

DOI: 10.1109/access.2026.3707611 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Access, 2026. Verfasst von K. Adi Narayana Reddy und 3 weiteren Autor:innen. Die Arbeit nennt 18 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Traditional approaches heavily rely on sentiment polarity classification. This approach groups opinions as “positive,” “negative,” or “neutral”; however, while effective for the implementation of a basic opinion mining strategy, this rudimentary classification fails to identify the large amounts of diverse, rich, and emotional data necessary for understanding the huge volume and variety of sentiments being expressed by human beings through large amounts of unstructured data. The growth of social media, online reviews, and digital communication has presented a need for advanced Context-Aware Emotion Intelligence Systems capable of detecting complex contextual emotional patterns in large heterogeneous data sources. This research proposes a scalable transformer-based emotion intelligence framework for contextual emotion classification using distributed Big Data processing. The framework integrates distributed preprocessing, transformer-based contextual representation learning, and emotion dependency modeling to classify text into six target emotional categories. The proposed framework utilizes distributed computing infrastructure and transformer-based contextual embeddings to efficiently process large-scale unstructured textual data and generate contextual emotion-aware sentiment intelligence. Unlike traditional polarity-based sentiment analysis systems, the suggested framework combines distributed Big Data processing with transformer-based contextual emotion modelling to develop a scalable and contextualized Emotion Intelligence Model. The emotion intelligence framework gets a macro-F1 score of 0.91 on the GoEmotions (58,009 instances) and SemEval emotion datasets; it outperforms emotion-aligned and polarity-based baseline models by about 4-6% and more than 10%, respectively. The distributed Spark-based implementation demonstrates scalable processing behaviour with substantial processing-time reduction under large-scale distributed workloads. The proposed framework formulates emotion prediction as a dominant-emotion single-label multi-class contextual classification problem while incorporating inter-emotion dependency modelling for enhanced contextual representation learning.

Zitieren

APA

Reddy, K. A. N., Ahmad, S., Chiranjeevi, P., & Supraja, T. (2026). Beyond Polarity: A Scalable Transformer-Based Emotion Intelligence Framework With Distributed Big Data Optimization. IEEE Access, 14, 101726-101745. https://doi.org/10.1109/access.2026.3707611

Vancouver

Reddy KAN, Ahmad S, Chiranjeevi P, Supraja T. Beyond Polarity: A Scalable Transformer-Based Emotion Intelligence Framework With Distributed Big Data Optimization. IEEE Access. 2026;14:101726-101745. doi:10.1109/access.2026.3707611

BibTeX
@article{Reddy2026,
  author = {Reddy, K. Adi Narayana and Ahmad, Shadab and Chiranjeevi, Pandi and Supraja, Thatikonda},
  title = {Beyond Polarity: A Scalable Transformer-Based Emotion Intelligence Framework With Distributed Big Data Optimization},
  journal = {IEEE Access},
  year = {2026},
  volume = {14},
  pages = {101726--101745},
  doi = {10.1109/access.2026.3707611}
}
Kostenlos starten →