Abdulrahman Alothaim

Progress in Arabic opinion summarization has been limited by the absence of standardized evaluation frameworks, despite Arabic being the fifth most widely spoken language globally. We present ArabicOpinion, the first large-scale, human-evaluated benchmark and automated evaluation framework for Arabic opinion summarization. The framework introduces entity-aware evaluation that assesses named entities alongside opinions, addressing a key gap in existing work. It includes evaluation metrics covering five dimensions: relevance, faithfulness, sentiment preservation, abstractiveness, and genericity, with a bidirectional multi-granular architecture that measures the first two independently. We benchmark GPT-4o, Claude, and JAIS. The automated metrics show strong alignment with human judgments. Opinion-level Matryoshka embeddings perform best for relevance and faithfulness, while AraBERT-Restaurant-Sentiment achieves the highest sentiment preservation accuracy. These results provide a standardized foundation for evaluating Arabic opinion summarization and offer insights into the performance of large language models on Arabic content.

Aspect-based sentiment analysis identifies fine-grained opinions in user reviews. However, most systems simply predict sentiment labels without revealing the text spans that justify those labels. Existing approaches extract minimal opinion terms, which cannot verify predictions, explain reviewer attitudes, or support downstream applications such as aspect classification, opinion summarization, and entity-level analysis. Addressing this, we introduce atomic opinion quote extraction, a formulation that extracts complete, minimal, self-contained evaluative spans exactly as they appear in the original text, structured as (aspect category, sentiment polarity, opinion quote) triplets.We created a large-scale dataset of 14,783 real-world restaurant reviews written in Arabic, spanning diverse dialects and annotated with 63,620 opinion triplets covering 29 aspect categories. This represents the first benchmark for atomic opinion quote extraction in Arabic.We further proposed a hybrid token–span architecture that combines sequence labeling with a boundary-aware span classifier to extract the atomic opinion quotes with precise boundary detection. Experimental results show that, on this challenging extraction task, the proposed model achieves 65.34% exact-match F1 and 81.67% F1 at a 50% overlap threshold, consistently outperforming strong token- and span-based baselines with statistically significant improvements. Combined, this task, dataset, and model advance Arabic sentiment analysis beyond mere label prediction toward explainable, fine-grained, and evidence-based opinion understanding, with broad applicability to downstream Arabic natural language processing tasks.

The dynamic shifts in educational settings have led scholars to explore online social networks (OSNs) as emergent environment for learning communities. Recognized for their effectiveness in fostering lifelong learning, these online groups surpass traditional educational methods in scope and impact. However, they require meticulous management to maintain efficacy and facilitate knowledge sharing. Addressing this need, the current study introduces the objective of identifying and defining learning leaders to aid in the management of these learning communities. To achieve this, two innovative metrics have been developed, rooted in the theoretical framework of Community of Practice (CoP) and the principles of information entropy: the Learning Leader Metric (LLM) and the Weighted Learning Leader Metric (WLLM). The LLM evaluates three key areas: Learning Engagement, Expertise, and Domain Relevance, forming a comprehensive tripartite framework. In contrast, the WLLM enhances this evaluation by applying information entropy to assign varying weights to these dimensions, thus refining the assessment process while preserving computational efficiency. These metrics operate in an unsupervised manner, addressing the challenge of the absence of annotated data which is a common hurdle in supervised learning scenarios, particularly in information retrieval tasks. Tested against five established expert finding models, the empirical results of this study confirm that both LLM and WLLM demonstrate superior effectiveness in accurately identifying highly relevant experts compared to other models considered in the analysis.

One of the main tasks in the field of natural language processing (NLP) is the analysis of affective states (sentiment and emotional) based on written text, and attempts have improved dramatically in recent years. However, in studies on the Arabic language, machine learning or deep learning algorithms were utilised to analyse sentiment and emotion more often than current pre-trained language models. Additionally, further pre-training the language model on specific tasks (i.e., within-task and cross-task adaptation) has not yet been investigated for Arabic in general, and for the sentiment and emotion task in particular. In this paper, we adapt a BERT-based Arabic pretrained language model for the sentiment and emotion tasks by further pre-training it on a sentiment and emotion corpus. Hence, we developed five new Arabic models: QST, QSR, QSRT, QE3, and QE6. Five sentiment and two emotion datasets spanning both small- and large-resource settings were used to evaluate the developed models. The adaptation approaches significantly enhanced the performance of seven Arabic sentiment and emotion datasets. The developed models showed excellent improvements over the sentiment and emotion datasets, which ranged from 0.15–4.71%.

The Arabic language is a complex language with little resources; therefore, its limitations create a challenge to produce accurate text classification tasks such as sentiment analysis. The main goal of sentiment analysis is to determine the overall orientation of a given text in terms of whether it is positive, negative, or neutral. Recently, language models have shown great results in promoting the accuracy of text classification in English. The models are pre-trained on a large dataset and then fine-tuned on the downstream tasks. Particularly, XLNet has achieved state-of-the-art results for diverse natural language processing (NLP) tasks in English. In this paper, we hypothesize that such parallel success can be achieved in Arabic. The paper aims to support this hypothesis by producing the first XLNet-based language model in Arabic called AraXLNet, demonstrating its use in Arabic sentiment analysis in order to improve the prediction accuracy of such tasks. The results showed that the proposed model, AraXLNet, with Farasa segmenter achieved an accuracy results of 94.78%, 93.01%, and 85.77% in sentiment analysis task for Arabic using multiple benchmark datasets. This result outperformed AraBERT that obtained 84.65%, 92.13%, and 85.05% on the same datasets, respectively. The improved accuracy of the proposed model was evident using multiple benchmark datasets, thus offering promising advancement in the Arabic text classification tasks.

DevOps (development and operations) is a set of collaborative practices that automate delivery of new software updates with the aim to reduce the development life cycle and produce quality software products. Software organizations face several barriers while adopting DevOps practices as the integration of development and operation teams requires merger of different processes, tools, and skill sets. This study aims to develop a prioritization-based framework of the DevOps best practices based on evidence collected from industry experts. To attain the study aims, firstly, a systematic literature review was conducted to identify DevOps best practices reported in the literature. Next, a questionnaire survey study was conducted to receive insight from industry practitioners for the identified best practice. Finally, the fuzzy-AHP technique was applied to prioritize the best practices concerning to the significance for DevOps process. We believe that the identified best practices, their categorization and fuzzy-AHP based framework will help industry experts to revise and improve their strategies to make the DevOps process sustainable.

Binary images have found its place in many applications, such as digital forensics involving legal documents, authentication of images, digital books, contracts, and text recognition. Modern digital forensics applications involve binary image processing as part of data hiding techniques for ownership protection, copyright control, and authentication of digital media. Whether in image forensics, health, or other fields, such transformations are often implemented in high-level languages without formal foundations. The lack of formal foundation questions the reliability of the image processing techniques and hence the forensic results loose their legal significance. Furthermore, counter-forensics can impede or mislead the forensic analysis of the digital images. To ensure that any image transformation meet high standards of safety and reliability, more rigorous methods should be applied to image processing applications. To verify the reliability of these transformations, we propose to use formal methods based on theorem proving that can fulfil high standards of safety. To formally investigate binary image processing, in this paper, a reversible formal model of the binary images is defined in the Proof Assistant Coq. Multiple image transformation methods are formalized and their reliability properties are proved. To analyse real-life RGB images, a prototype translator is developed that reads RGB images and translate them to Coq definitions. As the formal definitions and proof scripts can be validated automatically by the computer, this raises the reliability and legal significance of the image forensic applications.

Technology affects almost every aspect of life and is constantly changing. Digital communication technology has made it easier and faster to connect people all over the world. Digital technology is used in varies fields, including business, industries, companies, and educational institutions. There are various benefits of technology; it is also associated with a number of risks and dangerous threats known as cybercrimes. Cybercrime is a criminal activity that targets digital technology, like a computer, a computer network, or a mobile device. Cybersecurity is the way we reduce the risk of becoming a victim of cybercrime. Cybersecurity is the process of defending against cyberattacks. By using these concepts, we investigated the interval-valued complex T-spherical fuzzy relations (IVCT-spherical-FRs) introduced in this paper. We studied the relationships between different types of cybersecurity and the sources of cyberattacks. Furthermore, the Hasse diagram for the interval-valued complex T-spherical partial order set and relation is developed. The concepts of Hasse diagram are being used to examine various cybersecurity techniques and practices. The most effective method is identified using the features of Hasse diagrams. Finally, comparison tests are used to demonstrate the benefits of the proposed methods.