A FRAMEWORK FOR ARABIC SENTIMENT ANALYSIS USING MACHINE LEARNING CLASSIFIERS

International audienceIn recent years, the use of Internet and online comments, expressed in natural language text, have increased significantly. However, it is difficult for humans to read all these comments and classify them appropriately. Consequently, an automatic approach is required to classify the unstructured data. In this paper, we propose a framework for Arabic language comprising of three steps: pre-processing, feature extraction and machine learning classification. The main aim of the proposed framework is to exploit the combination of different Arabic linguistic features. We evaluate the framework using two benchmark Arabic tweets datasets (ASTD, ATA), which enable sentiment polarity detection in general Arabic and Jordanian dialects. Comparative simulation results show that machine learning classifiers such as Support Vector Machine (SVM), Naive Bayes, MultiLayer Perceptron (MLP) and Logistic Regression-based produce the best performance by using a combination of n-gram features from Arabic tweets datasets. Finally, we evaluate the performance of our proposed framework using an Ensemble classifier approach, with promising results