Design and implementation of classical literature sentiment analysis system based on ensemble learning and graph neural network

• Proposes an innovative sentiment analysis system combining graph neural networks and ensemble learning to achieve refined analysis of classical literature, surpassing traditional methods' limitations. • Constructs a large corpus covering classics from various eras, using graph neural network technology to extract deep semantic features and internal structural connections of texts. • Employs ensemble learning strategies to integrate multiple high-performance GNN models, effectively handling long-range dependencies and context information, achieving a sentiment classification accuracy of 91.5%. • Optimizes model performance through ensemble learning methods, using bagging and boosting to significantly reduce false alarm rates and improve overall sentiment recognition accuracy to 93.7%. Classical literary works have attracted extensive attention in modern society because of their unique cultural memory and aesthetic value. However, due to the long history and evolution of language, how to accurately grasp its connotation, especially its emotional color, has always been a dilemma for researchers. This study is committed to the design and implementation of a classical literature sentiment analysis system based on ensemble learning and graph neural network, with the goal of breaking through the limitations of traditional methods and realizing the refined analysis of classical literature sentiment tendencies. By constructing a large-scale corpus covering classics from different eras, this study lays a solid data foundation for model training. Graph neural network technology is innovatively applied to sentiment analysis in classical literature, and the graph structure composed of lexical nodes and semantic edges is used to capture the deep semantic and structural connections of texts. At the same time, bagging and boosting ensemble learning strategies are introduced to optimize the performance of multiple GNN models and form a more robust decision set. Experimental results show that compared with traditional methods, the graph neural network has an accuracy of 91.5% for sentiment classification, and the ensemble learning further reduces the false positive rate, improving the overall emotion recognition accuracy of the system to 93.7%, providing an efficient and accurate innovative solution for sentiment analysis of classical literature.