Performance analysis using deep learning approaches for sentiment classification of tweets

Abstract

Social media refers to a collection of websites and programs that enable users to express and disseminate their opinions among diverse communities. For instance, popular social media platforms include Facebook, Instagram, and X (Twitter). Instantaneous communication, text, video, and idea sharing is possible with people all over the world. However, overuse can have detrimental effects like mental health issues, decreased productivity, and cyberbullying. In the end, social media is a great tool for networking, self-expression, and international connection when utilized appropriately. A text mining technique called sentiment analysis is used to identify and categorize the emotional tone of textual material. Predicting a sentence's sentiment—whether good, negative, or neutral—assists companies in improving the quality of their services through feedback, reviews, and social media comments. This research article aims to sort of recurrent neural network called a long short-term memory (LSTM) is specifically made to keep the neural network output for a particular input from either exploding or fading as it cycles through the feedback loops. An embedding layer, a single LSTM layer, and a dense layer at the end make up this architecture. To reduce over-fitting, dropout methods are used in between the LSTM layers.