A Personalized Markov Clustering and Deep Learning Approach for Arabic Text Categorization
Vasu Jindal · 2016
Text categorization has become a key research field in the NLP community. However, most works in this area are focused on Western languages ignoring other Semitic languages like Arabic. These languages are of immense political and social importance necessitating robust categorization techniques. In this paper, we present a novel three-stage technique to efficiently classify Arabic documents into different categories based on the words they contain. We leverage the significance of root-words in Arabic and incorporate a combination of Markov clustering and Deep Belief Networks to classify Arabic words into separate groups (clusters). Our approach is tested on two public datasets giving a F-Measure of 91.02%.