Arabic Privacy Policy Corpus and Classification

Abstract

Privacy policies are the main sources of information for Internet users on how organizations gather, use, and share personal data. However, most consumers find it difficult to read and become unaware of the content. To simplify it for consumers, an automatic technique verifies whether websites and applications comply with the established privacy and data protection laws in companies worldwide. The implementation of the Personal Data Protection Law (PDPL) in the Kingdom of Saudi Arabia (KSA) was effective on March 23, 2022. To the best of our knowledge, checking the compliance of Saudi websites with the PDPL has not yet been addressed. Hence, this project aims to automatically analyze the compliance of KSA privacy policies with PDPL using Machine Learning and Deep Learning techniques. We manually extracted and annotated a dataset of privacy policies from 1000 Saudi websites belonging to seven sectors. The Logistic Regression technique was used as the baseline model to train the dataset along with other classifiers, including Naïve Bayes, Support Vector Machine, Long Short-Term Memory (LSTM), Feed-Forward Neural Network (FFNN), ARABERT, MARBERT, and CamelBERT. The results show that both MARBERT and CamelBERT models achieved the best performance with a micro-average F1 score of 0.93, compared to the baseline approach and other models.