View article

The inherent morphological complexity of languages such as Arabic entails the exploration of language traits that could be valuable to the task of detecting and classifying sentiment within text. This paper investigates the relevance of using the roots of words as input features into a sentiment analysis system under two distinct domains, in order to tailor the task more suitably to morphologically-rich languages such as Arabic. Different wordrooting solutions are employed in conjunction with a basic sentiment classifier, in order to demonstrate the potential of mapping Arabic words to basic roots for a language-specific development to the sentiment analysis task, showing a noteworthy improvement to baseline performance.