Title: E-government system for racial discrimination detection in Arabic social media: integrating AraBERT and ktrain in a multimodal learning framework

Authors: Tarek Kanan; Ashraf Almhirat; Tariq Samarah; Ghassan Kanaan; Omar AlAzzam

Addresses: Faculty of Sciences and IT, Al Zaytoonah University of Jordan, Amman, Jordan ' Faculty of Sciences and IT, Al Zaytoonah University of Jordan, Amman, Jordan ' Electronic Marketing and Social Media, Zarqa University, Zarqa, Jordan ' Faculty of Sciences and IT, Al Zaytoonah University of Jordan, Amman, Jordan ' Department of Computing, Informatics and Data Science, Saint Cloud State University, Saint Cloud, MN, USA

Abstract: This study presents an advanced NLP and multimodal learning framework to detect racial discrimination in Arabic social media. We curated a dataset of 10,319 Facebook and Twitter posts, applying preprocessing steps like normalisation, stopword removal, and stemming. The methodology integrates machine learning (SVM) and deep learning (RNN/LSTM), achieving high F1 scores. The best performance came from combining AraBERT and ktrain, with F1 scores of 90% on Twitter and 91.46% on Facebook. Our results demonstrate the effectiveness of this approach in improving classification accuracy in addition contributing to the development of more secure and comprehensive digital environments for Arabic speakers.

Keywords: multimodal learning; NLP; natural language processing; racial discrimination detection; deep learning; Arabic BERT; AraBERT integration; ktrain integration; social media analysis; text classification.

DOI: 10.1504/IJEG.2025.151217

International Journal of Electronic Governance, 2025 Vol.17 No.3, pp.284 - 300

Received: 19 Sep 2024
Accepted: 04 Mar 2025
Published online: 19 Jan 2026 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article