KanuriSenti: A novel dataset for sentiment analysis in the under-resourced Kanuri language

This paper presents KanuriSenti, a novel sentiment analysis dataset developed for Kanuri, a major yet under-resourced language spoken across Nigeria and the Lake Chad region. The dataset addresses a critical gap in Natural Language Processing by providing structured and sentiment-annotated Kanuri data, which has been largely absent from existing resources. KanuriSenti consists of a lexicon-based dataset containing over 10,000 entries labeled with sentiment polarity (positive, negative, neutral), and an affective E-ANEW-style dataset annotated across Valence, Arousal, and Dominance dimensions by native Kanuri speakers. Annotation consistency and reliability was validated using Cohen's Kappa, and lexical richness was measured using the Type-Token Ratio, confirming our dataset's suitability for sentiment-related tasks. While the dataset is designed specifically for sentiment analysis, its cultural and linguistic authenticity makes it a valuable benchmark for evaluating sentiment models in low-resource language settings and advancing equitable language technology in Africa.