Arabic Tweets-Based Sentiment Analysis Using Optuna-Tuned Bidirectional Gated Recurrent Unit Classifier
P Ashwini Kumari, Komuravelly Sudheer Kumar, Rajeshwari Patil, Myasar Mundher Adnan, Rajireddy Soujanya · 2024
Arabic sentiment analysis (ASA) is a dynamic field in Natural Language Processing (NLP) that applies Machine Learning (ML) techniques to understand the complex emotional tones that are present in Arabic text. By using ML techniques for sentiment analysis in Arabic, it can gain a deeper insight into human interactions within the verbal context. ASA is quite challenging due to its structural diversity and dialects. The proposed method uses a Bi-GRU classifier and optuna optimizer to enhance the ability to understand and analyze sequential data. The approach involves using a labeled Twitter dataset, pre-processing techniques, feature extraction, optuna optimization, and Bi-directional Gated Recurrent Unit (Bi-GRU) as a classifier. The results of this paper clarify the efficiency of DL models in accurately classifying sentiments in Arabic text, as illustrated by performance metrics such as 93.86® precision, 93.56® recall, 94.08® accuracy, and 93.88 F1 measure which is greater than the other existing methods like stacking logistic regression, Modified Switch Transformer (MST) and Bidirectional Long Short-Term Memory+ Dependency-based rules word embedding.