LSEL: A lightweight deep learning model for social-emotional classification of classical music

• We present the first SEL-oriented music analysis framework developed using an expert annotated SEM dataset grounded in established SEL theory. This framework enables the classification of three higher-order social-emotional competencies (PS, EPT, Outlook) through psychologically interpretable labels. • LSEL adopts a task-specialized lightweight CNN that preserves full 2D time–frequency structure and employs a streamlined activation and normalization strategy tailored to SEL semantics, resulting in only 6.19 × 10 5 parameters while retaining strong representational capacity. • LSEL demonstrated consistently superior performance under a unified evaluation protocol, outperforming five representative SOTA models. While these comparison models achieved accuracies between 64% and 82% on the SEM dataset, LSEL reached 96.55% accuracy and 99.29% mAP, indicating a substantial performance advantage. Social-emotional learning (SEL) plays a crucial role in special education, yet current assessment approaches rely heavily on subjective teacher observation, which can be time-consuming and difficult to standardize. Music provides a meaningful medium for evaluating emotional competencies, creating an opportunity for artificial intelligence to support more objective and scalable SEL assessment. We propose a lightweight social-emotional music classification model, termed LSEL, designed to identify three SEL-related competencies: Empathetic Perspective-Taking, Outlook, and Problem-Solving. LSEL utilizes 40×128 mel-frequency cepstral coefficient as input to capture core spectral–temporal characteristics relevant to SEL perception. Moreover, we provided an open-source SEM dataset for domain experts, utilizing 591 samples, which consisted of 194 Empathetic, 214 Outlook, and 183 Perspective-Taking samples, to analyze LSEL performance. LSEL reaching an average accuracy of 96.55% and mAP of 99.29% across experiments. With Cohen’s κ averaging 94.32% and R² reaching 94.15%, indicating high consistency with ground-truth. Per-category accuracies were similarly stable, including 96.95% for Empathetic Perspective-Taking, 95.16% for Outlook, and 95.36% for Problem-Solving. The lightweight LSEL framework offers an effective and robust solution for social-emotional music classification, supporting objective SEL assessment in educational contexts. The release of the SEM dataset further contributes to a valuable resource for advancing AI-assisted SEL research.