Aspect-Based sentiment analysis of Hausa language

Abstract

Aspect-Based Sentiment Analysis (ABSA) enables fine-grained understanding of tex tual data by identifying specific aspects of a target entity and determining the sen timent expressed toward each aspect. Despite significant advances in high-resource languages such as English, low-resource languages like Hausa remain underrepre sented, particularly in educational contexts. This study addresses this gap by devel oping an annotated Hausa-English ABSA dataset derived from an existing Hausa English sentiment dataset and enhanced for aspect-based analysis. The dataset was manually annotated with aspect categories and sentiment polarities to improve an alytical depth and applicability. Five key aspect categories were defined to reflect the educational domain: lecturer behavior, course content, course difficulty, teach ing quality, and teaching performance. Each instance was annotated with an aspect category and the corresponding sentiment polaritypositive, neutral, or negativepro viding a comprehensive resource for detailed sentiment interpretation. The objectives of this study include: (i) developing an annotated dataset for ABSA in the Hausa language using student comments; (ii) implementing machine learning models capable of extracting aspect terms and classifying their sentiment polarities; (iii) evaluating theperformanceofthemodelsinlow-resourcesettingsforABSAtasks; and (iv) contributing to the development of NLP tools and resources for underrepre sented African languages, with a focus on educational feedback analysis. Model per formance was assessed using standard metrics, including accuracy, precision, recall, andF1-score. Theresultshighlightchallengesinhandlingexplicitandimplicitaspect mentions inlow-resourcelanguagesandunderscoretheimportance ofrobustannota tion frameworks. ThisworkestablishesthefirststructuredHausaABSAdatasetinthe educational domain, providing a foundation for future multilingual and cross-lingual ABSAresearch in African languages.