What question did this study set out to answer?

The aim is to develop a robust transformer-based model for multimodal sentiment analysis in Arabic content.

March 7, 2026International Journal of Advanced Computer Science and Applications0 citationsOpen Access

Transformer-Based Multimodal Approach for Arabic Sentiment Analysis

A Transformer-Based Approach for Multimodal Arabic Sentiment Analysis

View Full Paper

Ask AI

Bookmark

Authors

ACAyoub Ben Cheikhi ENEL Habib NFAOUI

Discussion

Member takes

Overview

This research demonstrates enhanced sentiment prediction in Arabic via a tri-modal fusion model, indicating the potential for improved analysis integration.

Key Points

The aim is to develop a robust transformer-based model for multimodal sentiment analysis in Arabic content.
Fine-tune ViT for images, MarBERT for text, and HuBert for audio.
Implement early feature fusion of the three modalities.
Utilize classifiers for sentiment prediction.
Achieved state-of-the-art performance on the Ar-MuSA benchmark.
F1 score of 0.7756 and accuracy of 0.7759.
Surpassed unimodal and bimodal methods, highlighting the importance of tri-modal fusion.

Cite This Study

Cheikhi et al. (2026) studied this question.

synapsesocial.com/papers/69abc1b45af8044f7a4ea959 https://doi.org/10.14569/ijacsa.2026.0170290

View Full Paper

Ask AI

Bookmark

March 7, 2026International Journal of Advanced Computer Science and Applications0 citationsOpen Access

Transformer-Based Multimodal Approach for Arabic Sentiment Analysis

A Transformer-Based Approach for Multimodal Arabic Sentiment Analysis

View Full Paper

Ask AI

Bookmark

Authors

ACAyoub Ben Cheikhi ENEL Habib NFAOUI

Discussion

Member takes

Overview

This research demonstrates enhanced sentiment prediction in Arabic via a tri-modal fusion model, indicating the potential for improved analysis integration.

Key Points

The aim is to develop a robust transformer-based model for multimodal sentiment analysis in Arabic content.
Fine-tune ViT for images, MarBERT for text, and HuBert for audio.
Implement early feature fusion of the three modalities.
Utilize classifiers for sentiment prediction.
Achieved state-of-the-art performance on the Ar-MuSA benchmark.
F1 score of 0.7756 and accuracy of 0.7759.
Surpassed unimodal and bimodal methods, highlighting the importance of tri-modal fusion.

Cite This Study

Cheikhi et al. (2026) studied this question.

synapsesocial.com/papers/69abc1b45af8044f7a4ea959 https://doi.org/10.14569/ijacsa.2026.0170290

View Full Paper

Ask AI

Bookmark