Research Projects

Project Description
Opinion Mining for Arabic with Models of Semantics and Credibility  (funded by QNRF)

There is an increased interest in determining people’s opinions when, among many examples, seeking to buy products, sensing the public opinion on certain issues, or identifying trends. Governments and politicians are often interested in opinion (or sentiment) mining for defining policies and campaign strategies. Research on opinion mining in the Arab World is a timely subject as this region is going through important historical changes (the Arab Spring), which have been supported by the newfound freedom of expression that Arabs have access to online. To establish a solid foundation for effective opinion extraction, with special focus on Arabic text, our project has four specific solutions. First, we are building a large collection of resources (annotated text and lexica) for research on Arabic opinion mining. Second, we are exploring a large space of features for Arabic (orthographic, morphological, and syntactic) in the context of opinion mining. Third, we are developing and studying models of semantics and credibility for opinion mining in Arabic and English. The intuition behind these models is that they will allow us to capture human preconceived notions about the world. We are using these models as sources of additional features to automatically model opinions. Finally, we are combining the extracted opinion features to determine an effective opinion classifier, and developing advanced software prototypes to demonstrate and evaluate the developed solutions and techniques.