Opinion and sentiment polarity detection using supervised machine learning

This paper focuses on Opinion Mining (OM) and Sentiment Analysis (SA) for Arabic language. As there is a lack and size limitedness at lexicon level, we aim to build a new lexical resource following different methods, manually by extracting sentimental words from a selected dataset and semiautomatically by translating an English lexicon into Arabic. We also created a lemmatized version from an existing resource. These resources were subsequently used in the development of a polarity classifier. We begin this article by explaining the construction steps of these resources. Then, we present the supervised approach we developed to determine the polarity of the new data. The results of the tests carried out show the relevance of our choices.