Cross-Lingual Emotion Identification: A Collaborative Learning Approach for Cyber-Physical-Social Systems

Author First name, Last name, Institution

• Salman Ashraf, Ghulam Ishaq Khan Institute of Engineering Sciences and Technology
• Mousa Al-Kfairy, Zayed University, Abu Dhabi CampusFollow
• Babar Shah, Zayed University, Abu Dhabi CampusFollow
• Sajid Anwar, Institute of Management Sciences
• Zahid Halim, National Yunlin University of Science and Technology
• Bo Hsiao, National Yunlin University of Science and Technology

Document Type

Article

Source of Publication

IEEE Transactions on Computational Social Systems

Publication Date

1-1-2026

Abstract

Emotion identification from text is a critical area of natural language processing, with applications in understanding human sentiment and behavior. As communication increasingly occurs across linguistic and cultural boundaries, evaluating emotional content in diverse languages has become imperative. This study contributes to the advancement of collaborative learning and distributed intelligence in cyber-physical-social systems (CPSSs) by addressing challenges in cross-lingual emotion classification. We analyze publicly available data containing texts originally in English and translated into Spanish, German, and French. A novel machine learning framework is proposed, leveraging collaborative learning techniques with lexical and count-based textual features to model emotion across distributed linguistic systems. Results demonstrate that while emotions can be identified to some degree in all languages, performance decreases as the structural divergence of target languages from English increases, with French proving the most challenging. To validate these findings, we utilize the standardized public dataset international survey on emotion antecedents and reactions (ISEAR), which includes texts in the same four languages, and implement ensemble stacking and voting techniques to enhance distributed learning performance by combining multiple learner strengths. The obtained scores underline the complexities of integrating intelligent data-driven cyberspace with diverse linguistic and cultural domains. While representation learning and parameter optimization improved performance, fully addressing intrinsic barriers inhibiting direct transfer between structurally diverse languages remains an open challenge for the integration of human social intervention and physical distributed computing infrastructures.

DOI Link

10.1109/tcss.2026.3708567

ISSN

2329-924X

Publisher

Institute of Electrical and Electronics Engineers (IEEE)

Disciplines

Computer Sciences

Keywords

Cross-lingual emotion classification, distributed intelligence, ensemble learning, machine learning framework, multilingual systems

Scopus ID

105045275495

Recommended Citation

Ashraf, Salman; Al-Kfairy, Mousa; Shah, Babar; Anwar, Sajid; Halim, Zahid; and Hsiao, Bo, "Cross-Lingual Emotion Identification: A Collaborative Learning Approach for Cyber-Physical-Social Systems" (2026). All Works. 8227.
https://zuscholars.zu.ac.ae/works/8227

Indexed in Scopus

yes

Open Access

no