A Systematic Literature Review of Emotion Recognition in Conversation Using Natural Language Processing

Abstract:-Text-based emotion detection, especially in multi-speaker conversations, is now a very important area of research in Natural Language Processing (NLP). It has many useful uses in fields like customer service, healthcare, and social media analysis. This systematic literature review compiles the development of methodologies in this domain, mapping the transition from fundamental techniques to the contemporary state of the art. We start by looking at early lexicon-based and rule-based methods. These were fast to compute, but they had trouble with the subtleties of conversational language. The review then talks about how machine learning models like Support Vector Machines and Naive Bayes classifiers have made things better. These models used data-driven pattern recognition, but they were often limited by features that were made by hand. Deep learning is at the heart of modern emotion detection. Sequential models like LSTMs and transformer-based architectures like BERT are setting new performance standards by capturing temporal dependencies and bidirectional context. We also talk about hybrid and ensemble models that use more than one method to make predictions more accurate. This review also shows how important multimodal and contextual approaches are becoming. These approaches combine text with audio-visual cues and model how conversations work. Even though a lot of progress has been made, problems like dataset dependency, multilingual support, and ethical biases still exist. This paper concludes by delineating these challenges and charting future research trajectories, underscoring the necessity for more resilient, multimodal, and ethically-informed systems to enhance human-computer interaction.

---