CNN-based Emotion Classification Algorithm for Literary Texts in Secondary School Curriculum Tools

Emotional understanding is essential in literary education, enabling students to form a deep connection with texts and develop critical thinking skills. This study proposes EDUMOCNN, a new emotional classification structure that utilizes a 1D Multi-Scale Type Convolutional Network, integrated with an attention mechanism to identify emotional forms in aligned poetic texts within the secondary school curriculum. This model is trained and evaluated using Kaggle’s emotional-categorized poetry database, which comprises poems categorized into eight emotional types: anger, courage, fear, happiness, love, peace, sadness, and surprise. The use of multiple filters helps the model capture both the sample and extended emotional manifestations, while enlarged contexts allow for an expanded environmental model without increasing computational requirements. An attention layer is applied to highlight emotionally significant words, thereby improving category accuracy and educational value. The assessment results show that the standard works better than the CNN and LSTM basics in the EduEmoCNN, in terms of accuracy, Recall, and F1 Score, especially in distinguishing between common and subtle emotions in poetry literature. This structure promises to support teachers in emotionally based literary analysis and to coordinate with intelligent curriculum instruments to help students develop emotional literacy and intelligence. This research bridges the gap between AI and education by incorporating emotional intelligence into automated literary analysis.