MSA Repository "MSAR"

MSAR University's Digital Repository is a documentation and digitization of all university outcomes that are of effective value in the scientific and academic community and reflects the university's image, work, and effective contribution to society Through MSAR Digital Repository, the university managed to collect, store, archive and publish digital content - including documents, audio files, images and data sets - all in a safe place. MSAR is one of the strongest University Digital Repositories in Egypt and documented in the DSPACE community with its latest versions.

Communities in DSpace

Select a community to browse its collections.

Now showing 1 - 5 of 6

Recent Submissions

  • Thumbnail Image
    Item type: Item ,
    A comparative evaluation of machine translation vs. human translation for legal texts: A case study of translation between English and Arabic
    (Faculty of Modern Languages and Literatures, Adam Mickiewicz University, 2025-09-19) Noureldin Mohamed ABDELAAL; Islam AL SAWI
    This study examines the comparative accuracy and fluency of Neural Machine Translations (NMTs) and Language Model-based translations (LMBTs), represented by ChatGPT and Google Translate (GT), in legal texts translations. Texts from Farahaty's "Arabic-English-Arabic Legal Translation", sourced from primary texts cited in the book and translated by scholars such as Hatim, Shunnaq, Buckley, and Farahaty were used as benchmarks for human translation (HT). Sixteen diverse texts encompassing various legal discourse subgenres were selected for analysis, with all Arabic in-text examples transliterated using the Library of Congress (LOC) system. Qualitative analysis was conducted to assess the extent to which NMTs and LLMs match HT in accuracy and fluency. The study also investigated the similarities and differences between ChatGPT and GT in their translation outputs. Findings highlight HT's superiority in producing precise, stylistically appropriate translations, compared to the challenges faced by NMTs and LLMs in capturing legal terminology and subtle linguistic nuances. Despite variations, both ChatGPT and GT demonstrate efficiency and context sensitivity, suggesting their potential as valuable tools when coupled with human post-editing. The study concludes by advocating for a hybrid approach that leverages the strengths of automated translation systems and human expertise to enhance cross-linguistic legal communication.
  • Thumbnail Image
    Item type: Item ,
    Exchange Rate Policy, Transportation Costs, and Inflation Dynamics: Implications for Sustainable Economic Management in Egypt
    (World Association for Sustainable Development, 2026-05-20) Karim Soliman; Bandar Altubaishe; Doaa Mohamed Salman
    Purpose: This study focuses on determining the impact of exchange rate policy on transportation costs and inflation dynamics in Egypt, with implications on sustainable economic management. Design/methodology/approach: Using annual time-series data (2014-2023) from the Central Bank of Egypt, World Bank and International Monetary Fund (IMF), we use regression analysis techniques to estimate the effects of real interest rates, official exchange rates and exchange rate liberalisation on the Consumer Price Index for Transportation. Findings: Real interest rates significantly predict transportation costs (B=4.75, p<0.01). Exchange rate liberalisation has the biggest coefficient (beta=0.525) but it is not statistically significant, indicating the multicollinearity problem or low power. The official exchange rate does not matter a great deal. Descriptive statistics show high levels of volatility in all the indicators after devaluations. Originality/value: This study advances the literature by isolating transportation costs as a distinct exchange rate transmission mechanism in Egypt and examining its sustainability implications. Practical implications: Policy-makers should co-ordinate exchange rate flexibility with monetary policy, diversify energy inputs, and strengthen social safety nets. Research limitations: Small sample size and single-country focus limit generalisability.
  • Thumbnail Image
    Item type: Item ,
    FRunetm: Enhancing Biomedical Nucleus Segmentation with CBAM-SE Attention
    (Springer International Publishing AG, 2026-05-01) Menna Elgabry; Ali Hamdi; Mohamed Saiedur
    Nucleus segmentation is fundamental for interpreting biomedical images, which further enables the analysis of cell structures and disease recognition. In this research, we pro pose Fourier Residual UNet m (FRunetm), an enhanced version of the Fourier Residual UNet (FRunet) architecture enhanced with the Convolutional Block Attention Module Squeeze and excitation attention (CBAM-SE) attention mechanism, for improving segmentation precision on complex biomedical images. The FRunetm model was compared with three other deep learning models, which are Classic U-Net, SE U-Net, FRunet with a classic attention mechanism. All models were evaluated based on a spectrum of performance metrics, including accuracy, Dice coefficient, F1 score, loss, mean intersect over union (IoU), precision, and recall. Results show a marked segmentation performance improvement with the U-Net based architecture that employs the sophisticated attention mechanism (CBAM-SE). The current work outlines the performance of several models and the potential impact on the segmentation of medical images.
  • Thumbnail Image
    Item type: Item ,
    Deep Learning-Based Multiclass Classification of Oral Lesions with Stratified Augmentation
    (Springer International Publishing AG, 2026-05-01) Joy Naoum; Revana Salama; Ali Hamdi
    Oral cancer is frequently diagnosed at later stages due to its similarity to other lesions. Existing research on computer-aided diagnosis has made progress using deep learning; however, most approaches remain limited by small, imbalanced datasets and a dependence on single-modality features, which restricts model generalization in real-world clinical settings. To address these limitations, this study proposes a novel data-augmentation–driven multimodal feature-fusion framework integrated within a (Vision Recognition)VR-assisted oral cancer recognition system. Our method combines extensive data-centric augmentation with fused clinical and image-based representations to enhance model robustness and reduce diagnostic ambiguity. Using a stratified training pipeline and an EfficientNetV2-B1 backbone, the system improves feature diversity, mitigates imbalance, and strengthens the learned multimodal embedding. Experimental evaluation demonstrates that the proposed framework achieves an overall accuracy of 82.57% on 2 classes, 65.13% on 3 classes, and 54.97% on 4 classes, outperforming traditional single-stream CNN models. These results highlight the effectiveness of multimodal feature fusion combined with strategic augmentation for reliable early oral cancer lesion recognition and serve as a foundation for immersive VR-based clinical decision-support tools.
  • Thumbnail Image
    Item type: Item ,
    Confidence-Credibility Aware Weighted Ensembles of Small LLMs Outperform Large LLMs in Emotion Detection
    (Springer International Publishing AG, 2026-05-01) Menna Elgabry; Ali Hamdi
    This paper introduces a confidence-weighted, credibility-aware ensemble framework for text-based emotion detection, inspired by Condorcet’s Jury Theorem (CJT). Unlike conventional ensembles that often rely on homogeneous architectures, our approach combines architecturally diverse small transformer-based large language models (sLLMs)—BERT, RoBERTa, DistilBERT, DeBERTa, and ELECTRA—each fully fine-tuned for emotion classification. To preserve error diversity, we minimize parameter convergence while taking advantage of the unique biases of each model. A dual-weighted voting mechanism integrates both global credibility (validation F1- score) and local confidence (instance-level probability) to dynamically weight model contributions. Experiments on the DAIR-AI dataset demonstrate that our credibility-confidence ensemble achieves a macro F1-score of 93.5%, surpassing state-of-the-art benchmarks and significantly outperforming large-scale LLMs, including Falcon, Mistral, Qwen, and Phi, even after task-specific Low-Rank Adaptation (LoRA). With only 595 M parameters in total, our small LLMs ensemble proves more parameter-efficient and robust than models up to 7B parameters, establishing that carefully designed ensembles of small, fine-tuned models can outperform much larger LLMs in specialized natural language processing (NLP) tasks such as emotion detection.