Sara Sweidan

Dr. Sara Sweidan

orcid

Research Focus & Metrics

Academic Career Length

9Years (2018-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
2 July 2025 in Neural Computing and Applications
Verified
Journal Article
Closed access
PDF file
Fatma M. TalaatFatma M. Talaat
Fatma M. Talaat
RERana Mohamed El-Balka
Rana Mohamed El-Balka
SSSara Sweidan
Sara Sweidan
SGSamah Adel Gamel
Samah Adel Gamel
AAAYA MUHAMMED ALZOGHBY
AYA MUHAMMED ALZOGHBY

13 December 2018 in Arabian Journal for Science and Engineering
Verified
Journal Article
Closed access
PDF file
SSSara Sweidan
Sara Sweidan
SEShaker EL-Sappagh
Shaker EL-Sappagh
HEHazem El-Bakry
Hazem El-Bakry
SSSahar Sabbeh
Sahar Sabbeh
FBFarid A. Badria
Farid A. Badria
KKKyung Sup Kwak
Kyung Sup Kwak

9 August 2024 in IEEE Access
Verified
Journal Article
Open access
MSMustafa Abdul Salam
Mustafa Abdul Salam
MAMohammed Aldawsari
Mohammed Aldawsari
MGMostafa Gamal
Mostafa Gamal
HHHesham F. A. Hamed
Hesham F. A. Hamed
SSSara Sweidan
Sara Sweidan
Arabic language processing poses significant challenges owing to the scarcity of resources and intricate nature of its linguistic structures. To address these issues, this paper introduces MSG-ATS, a groundbreaking approach to abstractive text summarization in Arabic. The primary problem is the difficulty in creating coherent and high-quality summaries due to the complex syntactic, semantic, and contextual features of the Arabic language. This challenge is compounded by the lack of extensive linguistic and effective models tailored for Arabic summarization. MSG-ATS aims to overcome these challenges by leveraging multi-level representations within semantic graphs. This novel approach enhances conventional text construction and embedding techniques, allowing for a more nuanced capture of syntactic, semantic, and contextual information. This comprehensive capture is crucial for generating summaries that are both coherent and contextually relevant. By utilizing a deep neural network, MSG-ATS effectively generates high-quality summaries. To validate MSG-ATS, rigorous evaluations are conducted, comparing its performance with word2vec, a foundational word embedding method. These evaluations are supported by a novel dataset specifically compiled for this study. The generated summaries undergo thorough assessment, including automated evaluation using the Rouge metric. The results are compelling, demonstrating MSG-ATS’s superior performance with a significant 42.4% improvement in precision, a notable 23.8% improvement in recall, and an impressive 38.3% overall advancement over the baseline model.

20 November 2024 in Computers in Biology and Medicine
Verified
Journal Article
Closed access
SSSara Sweidan
Sara Sweidan
SAS.S. Askar
S.S. Askar
Mohamed AbouhawwashMohamed Abouhawwash
Mohamed Abouhawwash
EBElsayed Badr
Elsayed Badr

15 December 2025 in Arabian Journal for Science and Engineering
Verified
Journal Article
Open access
PDF file
SSSara Sweidan
Sara Sweidan
ATAhmed Taha
Ahmed Taha
Diagnosing retinal diseases is a fundamental challenge in developing robust multi-disease classification systems due to the limited availability of datasets and inconsistent quality. Therefore, this study presents $$\text{DeepRetina}$$ DeepRetina , a framework addressing these challenges through dataset harmonization. Five distinct retinal image datasets were unified using systematic preprocessing, resulting in a consolidated dataset of 29,966 high-resolution fundus images across eight disease categories. Harmonization corrects variations in image quality, color, and lighting resulting from different imaging devices or conditions. Data harmonization enhances the model’s ability to generalize across diverse datasets by standardizing the color and texture properties of images. The study compares the performance of custom CNN, EfficientNetV2, and MobileNetV3Large architectures for multi-disease classification. EfficientNetV2 achieved the highest accuracy of 79% with a precision of 54%. The proposed methodology significantly advances the field by (1) establishing a robust approach for harmonizing heterogeneous datasets, (2) presenting a large-scale, unified dataset for future research, and (3) presenting a comparative analysis of deep learning architectures optimized for retinal disease classification. $$\text{DeepRetina}$$ DeepRetina lays the foundation for scalable and accurate automated retinal disease diagnosis, contributing to improved detection and classification in ophthalmology.

26 January 2026 in Journal of Big Data
Verified
Journal Article
Open access
PDF file
SSSara Sweidan
Sara Sweidan
NFNorhan A. Farouk
Norhan A. Farouk
Mohamed AbouhawwashMohamed Abouhawwash
Mohamed Abouhawwash
SASameh Askar
Sameh Askar
MTMohamed Taha
Mohamed Taha
Machine-generated text, particularly deepfake content, presents substantial challenges for social media platforms. Detecting such content is crucial to maintaining trust and preventing misinformation. This study addresses the identification of machine-generated textual content on social media platforms. Specifically, we focus on short texts (e.g., tweets) by creating an intelligent framework to combat disinformation using the fake news detection framework, which employs the DeBERTaV3 model, the TweepFake dataset, and the PHEME dataset. The framework analyzes tweets and social media posts to determine whether a human or a bot account (machine-generated text) created the text. The efficacy of the suggested model is evaluated against various deep learning frameworks, including BERT, RoBERTa, SVM-RBF, random forest, SVM, CNN, and LSTM. Findings reveal that the fake news detection performance accuracy is 97.12%, signifying that the proposed model demonstrates exceptional performance and high accuracy. Experimental findings indicate that the DeBERTa architecture’s design, along with data preprocessing techniques and embedding methods, facilitates efficient and effective tweet classification, enabling the identification of whether a tweet is generated by a human or a bot account. The scalability and computational efficiency of fake news detection are evaluated against those of other models.