Self-labeling Approach for Spam Review Detection in Arabic Texts

Today, people share their experiences and opinions about products and services through online reviews. These reviews significantly impact online marketing and help us understand various products and services accurately. However, some of these online reviews can be fake and mislead consumers. These fake reviews, known as spam, are posted to promote or discredit a product. Previous research on detecting spam reviews has primarily focused on English reviews, with little attention given to other languages. Despite the vast amount of data generated, detecting spam reviews in Arabic online sources is still a new topic. This study addresses this gap by proposing a method for labeling a large dataset of Arabic reviews using a self-labeling approach. By utilizing a small set of labeled data, the method employs Random Forest classifiers to classify a more large, unlabeled dataset. Our approach improves accuracy and efficiency in spam detection and yields high-quality data with reduced quantity. Experimental results underscore the effectiveness of our method in producing more accurate classifiers.