Abstract
Understanding how emotions are expressed across languages is vital forbuilding culturally-aware and inclusive NLP systems. However, emotionexpression in African languages is understudied, limiting the development ofeffective emotion detection tools in these languages. In this work, we presenta cross-linguistic analysis of emotion expression in 15 African languages. Weexamine four key dimensions of emotion representation: text length, sentimentpolarity, emotion co-occurrence, and intensity variations. Our findings revealdiverse language-specific patterns in emotional expression -- with Somali textstypically longer, while others like IsiZulu and Algerian Arabic show moreconcise emotional expression. We observe a higher prevalence of negativesentiment in several Nigerian languages compared to lower negativity inlanguages like IsiXhosa. Further, emotion co-occurrence analysis demonstratesstrong cross-linguistic associations between specific emotion pairs(anger-disgust, sadness-fear), suggesting universal psychological connections.Intensity distributions show multimodal patterns with significant variationsbetween language families; Bantu languages display similar yet distinctprofiles, while Afroasiatic languages and Nigerian Pidgin demonstrate widerintensity ranges. These findings highlight the need for language-specificapproaches to emotion detection while identifying opportunities for transferlearning across related languages.