Sentiment Analysis of Social Media Posts with Hybrid BERT Models

Abstract

The rapid expansion of social media platforms has led to a surge in user-generated content, making sentiment analysis a crucial area of study for businesses, policymakers, and researchers. Traditional sentiment analysis models, including rule-based and machine learning approaches, often struggle with capturing complex linguistic structures and contextual nuances. With the advent of deep learning, BERT has emerged as a powerful NLP model capable of understanding bidirectional context, but standalone BERT models can be resource-intensive and may overfit smaller datasets. To address these challenges, we propose a Hybrid BERT model that enhances contextual understanding and classification performance. This research aims to develop and evaluate a Hybrid BERT model that integrates Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) layers to improve sentiment classification accuracy while maintaining computational efficiency. We fine-tuned a BERT base model and incorporated CNN layers to extract local text features along with BiLSTM layers to capture long-range dependencies. The Social Media Sentiments Analysis Dataset, comprising labeled positive, negative, and neutral posts, was used for training and evaluation. The model was optimized using Adam with a learning rate of 2e-5 and batch size of 32, and evaluated using accuracy, precision, recall, and F1-score. The proposed Hybrid BERT model achieved a sentiment classification accuracy of 95.67%, outperforming conventional deep learning models and demonstrating superior contextual comprehension. The model effectively reduced misclassification in ambiguous sentiment cases, highlighting its robustness in real-world applications. This study underscores the effectiveness of a Hybrid BERT model for sentiment analysis, significantly enhancing performance through improved contextual understanding. The findings suggest that such an approach is well-suited for applications in brand monitoring, social media analytics, and opinion mining.