The Classification Model Sentiment Analysis of the Sudanese Dialect Used Into the Internet Service in Sudan

Abstract

The analysis of Arabic opinions has not received much attention, such as English, primarily due to the challenges facing the elaboration of the complex Arabic language and the lack of tools and resources available to extract Arab sentiments from the text. This task is exacerbated when dealing with the Sudanese colloquial dialect that does not adhere to the formal grammatical structure of modern Standard Arabic. This study aims to analyze the opinions of the Internet service in Sudan written in the Arabic language using the modern Standard Arabic and Sudanese Standard Arabic accent, which was conducted on 1048 Facebook comments on the Internet service. The researcher used natural language processing techniques to process the data. Two different classifiers were applied: support vector machine (SVM) and Naive Bayes NB to classify comments based on their polarity either positive or negative. Then, the work was evaluated by four different measures as follows: accuracy, recall, accuracy, and measurement F. The results showed that SVM achieved the best accuracy—measurement which was 86.5% while NB achieved accuracy, which was equivalent to 80%.