Sentiment Analysis of Tweets During the Riyadh Season Using Deep Learning

Social media platforms have transformed interpersonal communication, reshaping the dynamics of human interaction and information dissemination. Platform X serves as a medium for global real-time communication, enabling individuals to share thoughts and ideas. Its open accessibility makes it a valuable resource for researchers seeking domain-specific data. Saudi Arabia prioritizes the development of its tourism sector, recognizing its significance for economic diversification and national advancement, in alignment with the goals of the Vision 2030 initiative. This study evaluates attendees' sentiments about the Riyadh Season in Saudi Arabia, an entertainment event. Tweets are categorized into positive, negative, and neutral sentiments using sentiment analysis techniques, allowing a comprehensive understanding of their perceptions and emotions. We collected the dataset using the Application Programming Interface (API) of X during the Riyadh Seasons of 2020 and 2021. The dataset contains 33,300 tweets. Through rigorous comparative analysis, the study assesses the performance of deep learning models, including Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTM) networks, hybrid CNN–Bidirectional LSTM (CNN-BiLSTM) architectures, and BiLSTM models with attention mechanisms. Notably, the BiLSTM model with an attention layer and Arabic Bidirectional Encoder Representations from Transformers (AraBERT) word embeddings emerges as the top-performing model, achieving an accuracy rate of 83.37%. This underscores the efficacy of BiLSTM models in capturing nuanced sentiment patterns, highlighting the importance of attention mechanisms and advanced word embeddings in enhancing sentiment analysis performance. In conclusion, this study contributes to sentiment analysis by providing insights into the perceptions of the Riyadh Season in Saudi Arabia and demonstrating the effectiveness of deep learning models in analyzing sentiment-rich X data.