Using Machine Learning and TF-IDF for Sentiment Analysis in Moroccan Dialect an Analytical Methodology and Comparative Study

The Moroccan dialect is a linguistic area that presents special difficulties because of its complex morphology and wide range of influences. This study offers a novel technique to sentiment analysis in this dialect. Our work focuses on using machine learning methods in conjunction with Natural Language Processing (NLP) techniques, namely Term Frequency-Inverse Document Frequency (TF-IDF) for feature extraction, to effectively classify sentiment. Given the scarcity of resources and standardized forms in Moroccan dialect, conventional sentiment analysis methods are less effective. To address this, our methodology involves rigorous preprocessing steps, including normalization, tokenization, and stemming, ensuring the refinement of input data for the machine learning models. The study utilizes a dataset comprising Moroccan tweets, classified into positive and negative sentiments, to train and test the models. We use algorithms such as Decision Tree, Support Vector Machine, and Logistic Regression, and assess their performance using metrics like accuracy, precision, recall, and F-1 score. Our findings highlight the varying effectiveness of these models in handling sentiment analysis for a morphologically rich and unstructured language like Moroccan dialect. This research not only contributes to the field of sentiment analysis in under-represented languages but also opens avenues for further exploration using more advanced NLP tools and deep learning techniques. It underscores the potential and challenges of applying machine learning to dialect-specific sentiment analysis, providing valuable insights for future research in this domain.