A Comparative Study of Text-Based Emotion Detection Techniques for Emotion Recognition on Social Media Data

Nowadays, with the rise of the internet social media have emerged as a new platform that provides a large amount of user-generated content that can be analyzed or mined to detect the emotions of the users. However, advancement in technology allows users to share their views through audio/video, speech, and facial expressions but people mostly use text for communication over social platforms. Hence it is required to automatically recognize emotions from these posts and provide responses to users in today’s online world. Emotion Detection from text is a popular topic in Natural Language Processing (NLP) due to its vast applications in human-computer interaction, artificial intelligence, marketing, etc. Researchers from different domains have proposed algorithms for emotion detection from the text; still, no method gave a complete solution. Hence, this paper aims to explore several emotion detection techniques such as machine learning (ML), Deep learning (DL), and hybrid ensemble techniques to detect emotions from social media conversations. Then these different classifiers are analyzed, tested, and compared with each other based on different parameters namely F1-score, recall, precision, accuracy, and confusion matrix. The classifier that gets the highest score among all other classifiers is termed as best text-based emotion classifier.