Abstract
Abstract Understanding the emotional states of university students is essential for addressing academic performance, well-being, and institutional support. This paper presents EduAffect, a novel emotion-labeled dataset constructed from public social media posts and structured survey responses by University Students in Ghana. The dataset comprises 2423 entries annotated using 28 fine-grained emotion labels (27 emotions plus Neutral) based on the GoEmotions framework. Each entry is tagged with a single emotion label, enabling context aware emotional analysis. The data reflects culturally grounded expressions in English and English-Twi code-switching. A thorough preprocessing pipeline ensured text quality, anonymisation, and language consistency. We provide detailed insights into the label distribution, lexical characteristics, and real-world examples, supported by a comprehensive taxonomy and sentiment mapping. This dataset fills a critical gap in student-focused emotion resources and offers a valuable foundation for future research in affective computing, student mental health, and emotion-aware educational technologies.