Abstract
Social networking sites have become an integral part of everyday life, where people interact, cooperate and quarrel with each other. Social media also encourages them to express their opinions and share their comments about their lives' events or about the product they use. Opinions can be direct without any comparison (I like ABC phone) or they can be comparative (X-phone's camera is better than Y-phone). Comparative opinions are useful in many applications, e.g. marketing intelligence, product benchmarking, and e-commerce. The automatic mining of comparative opinions is an important text mining problem and an area of increasing interest for different languages. This paper focuses on identification of comparative sentence from non-comparative ones in Arabic texts. A corpus was developed consisting of YouTube comments. This paper describes research experiments that aimed to apply data/text mining algorithms, natural language processing and linguistic classification for Arabic comparative text discovery. The results of these experiments along with the analysis are also presented. The results were promising reaching to 91% accuracy for the detection of comparative opinions.