Arabic Sentiment Analysis Using a Levenshtein Distance Based Representation Approach

Sentiment Analysis is one of the applications of the Natural Language Processing field undergoing the fastest development, and naturally, its need to cover the maximum amount of languages grows as well, and the Arabic language and its diverse dialects do not make the exception. In this perspective, we proposed a text data representation model based on the Bag of Words representation and the Levenshtein Distance. We applied this method on a dataset made of Moroccan dialect comments, to detect their polarity using a deep neural network classifier and got an accuracy of 62%.