Employing Attention-based BERT-BiRNN-CNN for Twitter Sentiment Analysis A Sentiment Analysis and Deep Neural Networks Approach

With the rise of social media, particularly Twitter, textual data containing users’ opinions on various topics has significantly increased. So sentiment analysis, which involves detecting the polarity of emotions in texts, has also grown. It uses machine learning techniques to accurately interpret and categorize texts, and record public sentiments in different areas. Sentiment analysis improves services and increases user satisfaction. In recent research, deep learning has gained considerable attention in sentiment analysis. Previous sentiment analysis studies have struggled with inadequate performance, focusing on words separately and not sufficiently on contextual information. Our research emphasizes the surrounding context of words and background information. Notably, complex and deep architectures perform better in learning intricate human emotions. Our context-based model elevates previous models utilizing BERT with attention mechanisms alongside Dilated Convolutional Neural Networks. This model results in a broader window for capturing local patterns after bidirectional Recurrent Neural Network layers resulting in a satisfactory sentiment analysis system. The accuracy of this research has improved, demonstrating considerable performance compared to previous studies.