Automatic Classification of Arabic News and Column Articles using Machine Learning and Deep Learning Approaches
Hanen Himdi, Bayan Alotaibi, Dania Alsahafy, Gharam Alghamdi, Layan Alghamdi · International Journal of Engineering & Technology · 2023
News is a report of recent events that is distributed from its point of origin to receivers by journalists who work for commercial organizations. Columns, on the other hand, are writings presented in special sections on news platforms that provide content on a variety of subjects and are often written by individuals or small groups of authors. Usually, columns might present information in a more subjective manner compared to the objective constraints in news reports. However, a major concern is that many columns lack the news’s editorial oversight process presented in news production, which can have negative effects, such as including incorrect information. In some cases, the presence of column articles on news platforms is mistaken for a news article, causing credibility concerns. To tackle this problem, a wide range of extant studies conducted in the recent past offer suitable techniques for column article classification. However, there are no studies in Arabic for this purpose. In this study, we introduce the first Arabic column article dataset that includes more than 12k articles. Then, we compiled several classification models using machine learning and deep learning approaches. It was found that deep learning models, CNN-LSTM, trained by BERT achieved the highest accuracy, reaching 96.6%. Finally, we propose a web platform that can be used freely to classify news and column articles based on their textual content solely