THE PURPOSE OF DEEP LEARNING MODEL USING EMBEDDING TECHNIQUE IN ARABIC SENTIMENT ANALYSIS

Abstract

Social media, widely used by Internet users to express their opinions on a given topic, has become one of the main sources of information for analysts. Sentiment analysis (SA) is a growing area of research of natural language processing (NLP) and machine learning (ML) tools to identify and label opinion text. Sentiment analysis is an important task in fields related to data analysis and information mining. In this study, the books of the most famous Arab authors were read and each sentence was manually extracted and labeled. This research aimed to generate a new Classical Arabic dataset (CASAD). In addition, feature extraction from these datasets is generated using word embedding techniques equivalent to Word2vec, which can extract deep relationships representing features of formal Arabic languages. Some machine learning techniques, such as support vector machine (SVM), logistic regression (LR), naive bayes (NB), K-nearest neighbor (KNN), latent Dirichlet allocation (LDA), and classification tree and regression are used to evaluate the features for classical Arabic (CART). In addition, statistical techniques such as validation and reliability are used to evaluate the labels of this dataset. Finally, using six machine learning algorithms for 10-fold cross-validation, our tests evaluated the classification rate of the feature extraction matrix into two and three classes, and the results showed that the Logistic regression with Word2Vec was the most effective in predicting the occurrence of polarizing topics.