Arabic Opinion Mining. Addressing the problems of Unbalanced Datasets and Huge Dimensionality

      Аннотация: Opinion Mining is a research area that aims to the analysis of people’s opinions towards entities, individuals, events, topics and their properties. The task is technically challenging and practically very useful. For example, businesses are always interested in consumer opinions about their products and services. Also, consumers themselves are interested in the opinions of other users to make purchase decisions. Opinion Mining involves several tasks where the most interesting one is Sentiment Classification, which consists in classifying opinionated texts as either positive or negative. Though the Arabic language is an important language, it has not gained enough attention from the research community of opinion mining given its lack in web content and its challenging processing. In this thesis, we focus on the classification of Arabic opinions by the use of supervised learning techniques. Though text classification is a mature area, some challenges remain to be addressed like the problems of huge dimensionality and unbalanced data sets, which are further accentuated when applied on opinions. In this thesis, we aim to give some research solution to these problems.

Ключевые слова: Computer Science