UzEDU-ABSA: An Uzbek Educational Domain Dataset for Aspect-Based Sentiment Analysis

UzEDU-ABSA is a UTF-8 encoded Uzbek educational domain dataset for aspect-based sentiment analysis (ABSA). The dataset contains 7,500 annotated text-aspect instances in a single tab-separated file, full.tsv.

Each row includes an Uzbek text sample, aspect term, aspect category, normalized aspect label, opinion expression, sentiment label, numeric polarity score, sentiment intensity, context type, negation flag, sarcasm/irony flag, and writing-system label. The dataset covers 30 education-related aspect categories, including teacher competence, teaching clarity, teaching methodology, teacher attitude, course content, curriculum, practical training, assessment fairness, exam process, feedback, workload, schedule, attendance, LMS platform, technical support, classroom infrastructure, learning resources, administrative services, payment and contracts, scholarship support, student support, internship and career, learning outcomes, academic environment, inclusion and equality, safety, dormitory and campus services, language and communication quality, digital content quality, and overall satisfaction.

The sentiment labels are POS, NEG, NEU, and MIX. The dataset contains 2,577 negative, 2,517 positive, 1,235 mixed, and 1,171 neutral instances. Polarity scores range from -2 to 2. The dataset also includes sentiment intensity labels (past, oʻrta, kuchli), context type labels, binary negation annotations, binary sarcasm/irony annotations, and script labels for Latin, Cyrillic, and mixed Uzbek writing.

This dataset is intended for research and development in Uzbek natural language processing, especially aspect category classification, aspect term extraction, opinion expression extraction, sentiment classification, polarity scoring, negation detection, and sarcasm/irony detection in the educational domain.