N-Gram Based HASSANIYA Dialect Classification

This paper focuses on the HASSANIYA dialect, which consists of the same Arabic letters and is widely used locally as a Mauritania dialect. Despite other local dialects, it has also been studied in terms of sentiment analysis. Yet, it still needs further studies to understand how word roots could be extracted and provide clear directions for future research in the relevant axis of Natural Language Processing (NLP), as text classification by machine learning techniques is largely influenced by the affixes of words, which could change the meaning of words. In this work, we present a novel approach that aims to study the effect of the N-gram method on a HASSANIYA dataset containing around two thousand comments collected from the Facebook platform and evaluate how it influenced the text classification results after preprocessing them. Additionally, three machine learning techniques, such as Support Vector Machine (SVM), Random Forests (RF), and Logistic Regression (LR), were applied for evaluation. The experiments give satisfactory results, confirming that the unigram method performs better than the bigram method for all algorithms.