Research on Machine Learning-Based Sentiment Recognition System

Abstract

This study explores the design, implementation, and evaluation of a machine learning-based emotion recognition system. A modular system is designed, covering core modules such as data processing, feature extraction, and emotion classification. Multiple machine learning models were evaluated experimentally, with results showing that the BERT model performed the best. The study highlights the application value of emotion recognition in mental health assessment, such as the early identification of psychological issues through the analysis of social media texts. This research provides a comprehensive reference framework for emotion recognition systems, making a significant contribution to advancing the field and its application in mental health.