Soufiene Jaffali
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Soufiene Jaffali has a total of 14 co-authors. Here are the top co-authors by number of shared publications:
- Abdelmajid Ben Hamadou
University of Sfax
6 shared publications - Nesrine Khelifi
Department of Management Information Systems, College of Business and Economics, Qassim University
3 shared publications Salma Jamoussi
University of Sfax
3 shared publications- Salma Jamoussi
Higher Institute of Computer Science and Multimedia of Sfax, University of Sfax
3 shared publications - Kamel Smaïli
Université de Lorraine
2 shared publications - Abdelwaheb Trabelsi
College of Computation and Informatics, Saudi Electronic University, Saudi Arabia, Dammam
1 shared publication
Latest Publications
- 17 April 2026
- Published by Taylor & Francis in The Journal of the Textile Institute
- Vol. 117 (7) , 1846-1863
Genetic algorithms for quality improvement in fabric roll cutting problem: a real-world industry case study
- 17 April 2026
- Published by Taylor & Francis in The Journal of the Textile Institute
- Vol. 117 (7) , 1846-1863
Genetic algorithms for quality improvement in fabric roll cutting problem: a real-world industry case study
- Open Access
- 5 December 2025
- Published by Asociatia EduSoft in BRAIN. BROAD RESEARCH IN ARTIFICIAL INTELLIGENCE AND NEUROSCIENCE
- Vol. 16 (4) , 113-129
Robust Sentiment Analysis Through Bayesian Dropout-Enhanced RoBERTa-LSTM
Transformer models such as RoBERTa provide strong contextualised embeddings for sentiment analysis but lack inherent mechanisms to quantify predictive uncertainty and are prone to overfitting, particularly on noisy or imbalanced text data. This paper introduces RoBERTa-LSTM-Drop, a hybrid architecture that combines RoBERTa embeddings with bidirectional LSTM layers and integrates Bayesian Dropout to capture uncertainty through Monte Carlo sampling while acting as an effective regulariser. The model is evaluated on two complementary benchmark datasets: IMDb, representing long and structured reviews, and Sentiment140, representing short and informal tweets. Comparisons are made against traditional baselines such as Logistic Regression as well as a RoBERTa-LSTM without Bayesian Dropout. Results demonstrate that RoBERTa-LSTM-B.Drop achieves accuracy competitive with strong baselines while offering calibrated uncertainty estimates that enhance model reliability. The findings highlight a stability–performance trade-off: Bayesian Dropout can yield modest accuracy improvements but introduces variance across optimisers and learning rates. Practical mitigation strategies, including dropout-rate adjustment and gradient clipping, are discussed. Beyond raw performance, uncertainty estimates enable identification of low-confidence, error-prone predictions, supporting risk-aware deployment in real-world scenarios. The analysis further clarifies when uncertainty-aware hybrids are most advantageous—such as in longer, context-rich inputs—and where their gains are limited. The work contributes both an empirical study of uncertainty-aware sentiment analysis and reproducible design choices to inform future research.
- 5 December 2025
- Published by Asociatia EduSoft in BRAIN. BROAD RESEARCH IN ARTIFICIAL INTELLIGENCE AND NEUROSCIENCE
- Vol. 16 (4) , 113-129
Open AccessRobust Sentiment Analysis Through Bayesian Dropout-Enhanced RoBERTa-LSTM
Transformer models such as RoBERTa provide strong contextualised embeddings for sentiment analysis but lack inherent mechanisms to quantify predictive uncertainty and are prone to overfitting, particularly on noisy or imbalanced text data. This paper introduces RoBERTa-LSTM-Drop, a hybrid architecture that combines RoBERTa embeddings with bidirectional LSTM layers and integrates Bayesian Dropout to capture uncertainty through Monte Carlo sampling while acting as an effective regulariser. The model is evaluated on two complementary benchmark datasets: IMDb, representing long and structured reviews, and Sentiment140, representing short and informal tweets. Comparisons are made against traditional baselines such as Logistic Regression as well as a RoBERTa-LSTM without Bayesian Dropout. Results demonstrate that RoBERTa-LSTM-B.Drop achieves accuracy competitive with strong baselines while offering calibrated uncertainty estimates that enhance model reliability. The findings highlight a stability–performance trade-off: Bayesian Dropout can yield modest accuracy improvements but introduces variance across optimisers and learning rates. Practical mitigation strategies, including dropout-rate adjustment and gradient clipping, are discussed. Beyond raw performance, uncertainty estimates enable identification of low-confidence, error-prone predictions, supporting risk-aware deployment in real-world scenarios. The analysis further clarifies when uncertainty-aware hybrids are most advantageous—such as in longer, context-rich inputs—and where their gains are limited. The work contributes both an empirical study of uncertainty-aware sentiment analysis and reproducible design choices to inform future research.
- Open Access
- 29 July 2025
- Published by Asociatia EduSoft in BRAIN. BROAD RESEARCH IN ARTIFICIAL INTELLIGENCE AND NEUROSCIENCE
- Vol. 16 (2) , 285-295
Unravelling the Power of Sentiment: How Emotions Shape Online Engagement in Public Health Discourse
Social media platforms have evolved into powerful arenas for public discourse, significantly influencing opinions, behaviours, and policy decisions. This study explores the relationship between sentiment and user engagement in online conversations related to public health. Using advanced sentiment analysis techniques powered by the BERT model, we analyse how emotional tone affects interactions such as likes, replies, and shares. Our findings reveal that positive sentiments drive higher engagement in terms of likes and replies, while negative sentiments generate more shares, amplifying their reach.. Neutral sentiments, in contrast, exhibit lower engagement across all metrics. Statistical analyses confirm significant differences in engagement based on sentiment, underscoring the critical role emotions play in shaping online conversations. These insights offer valuable guidance for policymakers, health communicators, and digital strategists aiming to craft more effective messaging strategies. By understanding how sentiment influences engagement, organisations can better connect with audiences, foster meaningful discussions, and enhance the impact of public communication campaigns.
- 29 July 2025
- Published by Asociatia EduSoft in BRAIN. BROAD RESEARCH IN ARTIFICIAL INTELLIGENCE AND NEUROSCIENCE
- Vol. 16 (2) , 285-295
Open AccessUnravelling the Power of Sentiment: How Emotions Shape Online Engagement in Public Health Discourse
Social media platforms have evolved into powerful arenas for public discourse, significantly influencing opinions, behaviours, and policy decisions. This study explores the relationship between sentiment and user engagement in online conversations related to public health. Using advanced sentiment analysis techniques powered by the BERT model, we analyse how emotional tone affects interactions such as likes, replies, and shares. Our findings reveal that positive sentiments drive higher engagement in terms of likes and replies, while negative sentiments generate more shares, amplifying their reach.. Neutral sentiments, in contrast, exhibit lower engagement across all metrics. Statistical analyses confirm significant differences in engagement based on sentiment, underscoring the critical role emotions play in shaping online conversations. These insights offer valuable guidance for policymakers, health communicators, and digital strategists aiming to craft more effective messaging strategies. By understanding how sentiment influences engagement, organisations can better connect with audiences, foster meaningful discussions, and enhance the impact of public communication campaigns.
- 13 January 2025
- Published by Elsevier in Journal of Molecular Liquids
- Vol. 421
Statistical comparison of models and machine learning for predicting Arrhenius parameters in ternary liquid mixtures
- 13 January 2025
- Published by Elsevier in Journal of Molecular Liquids
- Vol. 421
Statistical comparison of models and machine learning for predicting Arrhenius parameters in ternary liquid mixtures
- 6 June 2020
- Published by Springer Nature in Social Network Analysis and Mining
- Vol. 10 (1) , 1-11
Like-tasted user groups to predict ratings in recommender systems
- 6 June 2020
- Published by Springer Nature in Social Network Analysis and Mining
- Vol. 10 (1) , 1-11
Like-tasted user groups to predict ratings in recommender systems
- Open Access
- 1 June 2016
- Published by Springer Nature
- p. 3-14
Grouping Like-Minded Users for Ratings’ Prediction
- 1 June 2016
- Published by Springer Nature
- p. 3-14
Open AccessGrouping Like-Minded Users for Ratings’ Prediction
- 1 December 2014
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- No. 23759232,p. 921-927
Clustering and Classification of Like-Minded People from their Tweets
Several challenges accompanied the growth of online social networks, such as grouping people with similar interest. Grouping like-minded people is of a high importance. Indeed, it leads to many applications like link prediction and friend or product suggestion, and explains various social phenomenon. In this paper, we present two methods of grouping like-minded people based on their textual posts. Compared to three baseline methods K-Means, LDA and the Scalable Multistage Clustering algorithm (SMSC), our algorithms achieves relative improvements on two corpora of tweets.
- 1 December 2014
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- No. 23759232,p. 921-927
Clustering and Classification of Like-Minded People from their Tweets
Several challenges accompanied the growth of online social networks, such as grouping people with similar interest. Grouping like-minded people is of a high importance. Indeed, it leads to many applications like link prediction and friend or product suggestion, and explains various social phenomenon. In this paper, we present two methods of grouping like-minded people based on their textual posts. Compared to three baseline methods K-Means, LDA and the Scalable Multistage Clustering algorithm (SMSC), our algorithms achieves relative improvements on two corpora of tweets.
- 1 January 2014
- Published by Springer Nature in Lecture Notes in Computer Science
- p. 83-93
Grouping Like-Minded Users Based on Text and Sentiment Analysis
- 1 January 2014
- Published by Springer Nature in Lecture Notes in Computer Science
- p. 83-93
Grouping Like-Minded Users Based on Text and Sentiment Analysis