Arabic Poetry Meter Categorization Using Machine Learning Based on Customized Feature Extraction

Text mining applications became important in various intelligent tasks. Text documents are the most materials that record many important procedures in various worldwide organizations and different people cultures. Text poetry is an important type of people culture and education domains media. Arabic text poems classification is a few experimented fields, however, it has an important presence and special influence. Both new and ancient Arabic poetry has the same unique approach for rhythmical harmony measure, which can be used for identifying Arabic poems types. Deep learning as a machine learning method has many distinctive achievements in many areas, as well as, text classification tasks. In this paper, Arabic poetry text is categorized. A customized feature selection is proposed, which is fused with a clustering technique for enhancing models efficiency. Deep learning has experimented alongside two popular machine learning techniques; support vector machine and decision tree. The proposed feature extraction method has achieved high accuracy with all three techniques. The results are better than many related works.