Arabic Learning Resource Recommendation Dataset and Source Code

This repository provides the dataset and source code associated with research on personalized recommendation of Arabic learning resources based on learner sentiment. The repository includes: Code: source code and scripts used for data processing, feature extraction, sentiment-related analysis, and recommendation experiments. Data: learner-resource interaction records, pseudonymized identifiers, rating-derived sentiment labels, behavioral and linguistic features, and aggregated book-level statistics. Synthetic data: additional synthetic variables used in the experimental framework. Variants: alternative data configurations and experimental variants. Technical files: configuration parameters, validation reports, and documentation. The dataset does not contain original BRAD review texts or Arabic textual content. Original identifiers have been replaced with sequential pseudonymous identifiers. The mapping between original and pseudonymous identifiers is not included in this repository. The resource is intended to support reproducible research in Arabic Natural Language Processing, Sentiment Analysis, Learning Analytics, and Recommender Systems. Data provenance: Some records and attributes originate from the BRAD dataset, while other features are derived or synthetically generated. The original source and applicable usage conditions must be acknowledged. Privacy: The released identifiers are pseudonymized. This does not guarantee complete anonymization, as indirect linkage with external information may remain possible. Please cite this dataset using its DOI when referring to or reusing the materials made available in this record.