Emotion Recognition and Analysis System (ERAS): Leveraging Advanced AI for real-time Emotion Detection in Social Media Content

This study outlines the development of an advanced system for detecting and analyzing emotions expressed in social media content using cutting-edge artificial intelligence techniques in a web application. The application system utilizes advanced natural-language processing (NLP) and machine-learning techniques to distinguish emotions precisely in text-based content, including posts, comments, and tweets. By employing sentiment analysis, machine learning models, and emotion recognition algorithms within the developed web application Emotion Recognition and Analysis System (ERAS), the application utilizes 28 distinct emotional categories and employs 9 machine learning algorithms, such as Logistic-Regression, Linear-SVC, BERT-Classifier, Decision-Tree Classifier, AdaBoost-Classifier, Zero-shot Classifier, RNN-Classifier, Random-Forest, and Naive-Bayes Classifier including Logistic-Regression, which achieved the highest accuracy of 89%. The application itself delivers an accuracy of 90% above through a custom model, which dominates the performance of the Logistic-Regression model in some emotions like anger, fear, joy, sadness and so on. To develop this system, datasets were sourced from Kaggle and Hugging Face, ensuring comprehensive emotional coverage and alignment with research objectives. The GoEmotions dataset and selected Hugging Face datasets were primarily used for final model training due to their broad range of emotional categories. The evaluation of performance was carried out using metrics like precision, re-call, and F1-score. This application facilitates real-time thought analysis, presenting valuable insights for marketing, mental health, and social research.