In the part of education, understanding student sentiment is crucial for enhancing learning experiences and academic outcomes. This paper proposes a novel approach for educational student sentiment classification using a Extended Generative Semi-Encoder Based Convolutional framework (EGSCF) model. The methodology encompasses several phases tailored to effectively harness both textual and visual information for sentiment analysis in educational settings. Initially EGSCF is trained on a dataset comprising text data associated with educational contexts and student interactions. This model is designed to extract features and identify components that complement textual data; thereby subsequently fusion-based classification is employed to integrate the outputs of the sentiment classification model, utilizing textual features, with the visual features extracted by the EGSCF model. This research intends to show that the EGSCF-based sentiment categorization system may improve teaching methods and create more enjoyable learning environments via extensive testing and analysis.