Potential Approach for Text-Based Emotion Detection Using NLP Coupled With Deep Learning of Sentiment Analysis

Sentiment analysis is a valuable method for gauging people’s sentiments and emotions towards various subjects. Within this realm, emotion detection stands out by predicting specific emotions rather than categorizing them broadly as positive, negative, or neutral. While previous research has predominantly focused on emotion recognition through speech and facial expressions, text-based emotion detection poses distinct challenges due to the absence of cues like tonal stress and facial expressions. Addressing these challenges, natural language processing (NLP) techniques have been employed in the past, including the keyword approach, lexicon-based approach, and machine learning approach. However, keyword- and lexicon-based strategies have limitations, particularly in handling semantic relations. In this study, a propose novel hybrid model that combines machine learning and deep learning, specifically utilizing a sequential neural network architecture. The model includes embedding layers, flattening, and dense layers. The effectiveness of our approach is demonstrated through training on a diverse dataset encompassing sentences, tweets, and dialogs. Remarkably, our model achieves a high accuracy of the validation data, underscoring its efficacy in text-based emotion detection without relying on existing content.