Textual Emotion Recognition: Advancements and Insights

In recent years, there has been a significant increase in attention toward emotion detection in text analysis, driven by its broad applications across marketing, political science, psychology, human-computer interaction, and artificial intelligence. This growing interest is primarily due to the critical role of textual expression as a repository of human emotions and sentiments. The development of sophisticated natural language processing (NLP) techniques has emphasized the importance of exploring emotion detection and recognition within textual data. By utilizing a wide range of sources, including social media content, microblogs, news articles, and customer feedback, text mining aims to reveal the underlying emotional currents within the text. However, existing models often struggle to capture the complicated emotional nuances woven into words. Addressing this challenge, the innovative semantic emotion neural network (SENN) architecture has been introduced. The SENN model marks a significant advancement, featuring two synergistic sub-networks: a bidirectional long short-term memory (BiLSTM) network that extracts contextual information and a convolutional neural network (CNN) that analyzes and extracts emotional features, highlighting the text's intrinsic emotional connections. The SENN model's performance has been thoroughly evaluated on widely used real-world datasets, benchmarked against Ekman's six fundamental emotions. Results demonstrated its superiority, showing that the SENN model excels in emotion recognition accuracy and quality in conjunction with additional techniques. It also holds potential for enhancement by incorporating more comprehensive emotional word embedding, suggesting a promising future for text-based emotion analysis. The proposed paper presents goals for detecting sentiment in text data and introduces a novel architecture that effectively captures the complexity of emotional nuances. We create an abstract model and compare three types of machine learning, where logistic regression is the best. We also create an interactive model that allows the user to input text and display sentiments using API. The results of the proposed study show that the logistic regression model demonstrates a commendable balance between training and testing accuracy and robust performance across key metrics such as recall, F1 score, and precision.