Arabic dialects classification using text mining techniques

Arabic language is one of the important languages in the world. It is spoken in many countries. However, every region in the Arabic world has its own dialect. The spoken dialects start to be written recently with the rise of usage of social media tools such as Facebook and Twitter. In this paper, we discuss the development of a classifiers for six main Arabic dialects that are Gulf, Iraqi, Shami, Moroccan and Sudanese dialect, and Egyptian dialects. We used 2000 tweets that belonged to one of the six Arabic dialects to be used as data to our classifiers. The classifier was implemented using rapid miner tool. We used decision tree, naive Bayes and rule-based (Ripper) classification algorithms for classification purpose. The classifiers from three algorithms were able to classify the tweets into one of six dialects with some error rate but the classifier study revealed that algorithms were able to pick the keywords that are the salient features of the different dialects.