Social Network Sentiment Analysis Using Hybrid Deep Learning Models

Featured Application The system presented in this study was aimed at business, organisation, government, and consumer areas, whereas a real-time application is needed for the classification of social network messages written in Spanish through sentiment semantics metadata. Our model improved the performance of other existing machine learning techniques by up to 20 percentage points. These high levels of accuracy are crucial for obtaining real-time ratings from thousands of people for the effective monitoring of social media discourse in decision-making and strategy implementations.Abstract The exponential growth in information on the Internet, particularly within social networks, highlights the importance of sentiment and opinion analysis. The intrinsic characteristics of the Spanish language coupled with the short length and lack of context of messages on social media pose a challenge for sentiment analysis in social networks. In this study, we present a hybrid deep learning model combining convolutional and long short-term memory layers to detect polarity levels in Twitter for the Spanish language. Our model significantly improved the accuracy of existing approaches by up to 20%, achieving accuracies of around 76% for three polarities (positive, negative, neutral) and 91% for two polarities (positive, negative).