Emotion Detection From Text Using Machine Learning Approach

Abstract

Emotion detection from text plays a crucial role in various applications, from sentiment analysis on social media to customer feedback analysis in businesses. It is challenging due to the variability and ambiguity of human languages and emotions, which are context-dependent in themselves. This project aims to evaluate and compare seven machine learning models for predicting emotions in textual data. Utilizing a dataset sourced from various websites, which comprises 40,000 data points categorized into six emotion classes (happiness, sadness, anger, etc.). The study systematically addresses data acquisition, preprocessing, feature engineering, model training, and evaluation. The results indicate that the Random Forest Classifier model outperforms the others, achieving an accuracy of 85%. It improves emotion detection methodologies by trying to overcome the challenges presented by imbalanced datasets, thus contributing well to more efficient and accurate systems. This project will help businesses enhance customer satisfaction by accurately analyzing feedback and enable society to better monitor mental health and understand public sentiment. This study aims to highlight the limitations of textual emotion detection and propose future research directions by examining the concepts and performance of various models, approaches, and methodologies.