Exploring Cultural Nuances in Emotion Perception Across 15 African Languages
Ibrahim Said Ahmad, Shiran Dudy, Tadesse Destaw Belay, Idris Abdulmumin, Seid Muhie Yimam, Shamsuddeen Hassan Muhammad, Kenneth Church · 2025
Read Exploring Cultural Nuances in Emotion Perception Across 15 African Languages by Ibrahim Said Ahmad, Shiran Dudy, Tadesse Destaw Belay, Idris Abdulmumin, Seid Muhie Yimam, Shamsuddeen Hassan Muhammad, Kenneth Church (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-19642.txt.
Introduction
Ibrahim Said Ahmad and co-authors compare how feelings are expressed in fifteen African languages, in a study from two thousand twenty-five. Emotion tools work poorly when the languages behind them are barely studied. The team checked text length, tone, how feelings pair up, and how strong feelings run.
Somali texts ran long, while some other languages kept feelings brief. Several Nigerian languages showed more negative tone than others. Anger with disgust, and sadness with fear, paired up across many languages. Strength patterns split by language family, with some groups showing wide ranges. The authors urge tools built for each language, plus shared learning among kin languages.
Catalog record
| CREATOR | Ibrahim Said Ahmad, Shiran Dudy, Tadesse Destaw Belay, Idris Abdulmumin, Seid Muhie Yimam, Shamsuddeen Hassan Muhammad, Kenneth Church |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 25 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-19642
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Exploring Cultural Nuances in Emotion Perception Across 15 African Languages Ibrahim Said Ahmad, Shiran Dudy, Tadesse Destaw Belay, Idris Abdulmumin, Seid Muhie Yimam, Shamsuddeen Hassan Muhammad, Kenneth Church Abstract Understanding how emotions are expressed across languages is vital for building culturally-aware and inclusive NLP systems. However, emotion expression in African languages is understudied, limiting the development of effective emotion detection tools in these languages. In this work, we present a cross-linguistic analysis of emotion expression in 15 African languages. We examine four key dimensions of emotion representation: text length, sentiment polarity, emotion co-occurrence, and intensity variations. Our findings reveal diverse language-specific patterns in emotional expression—with Somali texts typically longer, while others like IsiZulu and Algerian Arabic show more concise emotional expression. We observe a higher prevalence of negative sentiment in several Nigerian languages compared to lower negativity in languages like IsiXhosa. Further, emotion co-occurrence analysis demonstrates strong cross-linguistic associations between specific emotion pairs (anger-disgust, sadness-fear), suggesting universal psychological connections. Intensity distributions show multimodal patterns with significant variations between language families; Bantu languages display similar yet distinct profiles, while Afroasiatic languages and Nigerian Pidgin demonstrate wider intensity ranges. These findings highlight the need for language-specific approaches to emotion detection while identifying opportunities for transfer learning across related languages. 1 Introduction The expression and perception of emotions are fundamental aspects of human communication, deeply intertwined with language and culture Mesquita et al. 1997 ; Wierzbicka 1999 . As natural language processing (NLP) advances, understanding how emotions are conveyed across different languages has become increasingly important. This paper explores the intricate relationship between emotion intensity, co-occurrence, and textual expression in various languages, aiming to uncover both universal trends and language-specific nuances. Emotions play a crucial role in shaping interpersonal interactions, influencing not only how we communicate but also how we interpret messages Leersnyder et al. 2011 . Previous research has primarily focused on emotion recognition and sentiment analysis within single or multiple languages, often overlooking the rich diversity of emotional expression across linguistic boundaries Tatariya et al. 2024 ; Saravia et al. 2018 ; Havaldar et al. 2023 ; Demszky et al. 2020 . Our study seeks to bridge this gap by examining emotion intensity and co-occurrence patterns in a multilingual context, thereby contributing to a more holistic understanding of emotion in language. In this paper, we address the following research questions: 1. Is there a correlation between the length of text and the intensity or type of emotions expressed? Do certain languages tend to use longer or shorter texts to convey specific emotions? 2. How do sentiment polarity distributions vary across typologically diverse languages in social media text, and what factors contribute to the observed differences in negative sentiment prevalence? 3. What are the prevalent patterns of emotion co-occurrence in various languages, and how do these patterns reflect cultural norms or communication styles? 4. How does the average intensity of emotions such as joy, sadness, anger, and surprise vary across different languages? What cultural or linguistic factors might contribute to these differences? To answer these questions, we analyze a diverse dataset comprising text samples from fifteen (15) African languages, employing NLP techniques to measure emotion intensity and co-occurrence. Our findings reveal insights into the interplay between language, culture, and emotion, highlighting both shared and distinct characteristics across languages. Our research contributes to the broader field of computational linguistics and offers practical implications for multilingual NLP applications, such as sentiment analysis, machine translation, and cross-cultural communication tools. Furthermore, our study underscores the importance of considering cultural and linguistic diversity in the development of emotion-aware technologies. 2 Related works Language and culture are deeply intertwined, a relationship extensively studied in psycholinguistics Zhou et al. 2025 . Recent advancements in NLP have heightened the need to understand this intersection in social media and online texts, ensuring AI systems accurately represent cultural perspectives. Current research evaluates how LLMs reflect cultural diversity in language generation Ahmad et al. 2024 ; Liu et al. 2024 ; Koto et al. 2024 ; Liu et al. 2024 ; Dudy et al. 2024 ; Cao et al. 2023 . Other efforts focus on building resources for culturally aware systems, such as CultureBank, a knowledge base for culture-specific data Shi et al. 2024 . Emotion datasets across cultures, such as Muhammad et al. 2025b , Belay et al. 2025 and Demszky et al. 2020 , further enable training inclusive AI models. Recent work also adopts multimodal approaches, analyzing cultural emotions in both text and images Mohamed et al. 2024 ; Khanuja et al. 2024 . While existing works leverage social media data to study cultural emotions, comparative analyses across languages remain limited. For instance, studies often focus on developing datasets rather than systematically contrasting how emotions manifest in different linguistic contexts. Social media’s informal, code-switched, and multilingual nature offers a unique opportunity to investigate these dynamics, yet few works explicitly compare cross-cultural emotional patterns between languages, such as Arabic versus Hausa or Amharic versus Igbo. Bridging this gap could uncover nuanced cultural divergences and improve AI’s adaptability to global linguistic diversity. 3 Experiments 3.1 Data Our study utilizes existing datasets by Muhammad et al. 2025b ; Muhammad et al. 2025c and Belay et al. 2025 that encompass text samples from multiple languages. We focus on 15 African languages: Algerian Arabic (ary), Amharic (amh), Emakhuwa (vmw), Hausa (hau), Igbo (igo), IsiXhosa (xho), IsiZulu (zul), Kinyarwanda (kin), Mozambique Portuguese (ptmz), Nigerian Pidgin (pcm), Oromo (orm), Swahili (swa), Somali (som), Tigrinya (tir), and Yoruba (yor). The dataset consists of a collection of texts from two sources, social media posts and news articles headlines, ensuring a comprehensive representation of emotional expression in different contexts. 3.2 Methodology As the goal of this project was to study the relation of emotional expression across language and culture, in the next sections, we provide a description of how we assess these relations. Our analysis focused on the following questions: Are there language differences in emotional expression? Can we characterize languages (and their specific sources) by their length? is emotion co-occurrence unique or similar across languages? We analyzed different key dimensions of emotional expression, such as text length distribution, emotion co-occurrence in the texts, and emotion intensity analysis. 4 Results 4.1 Text Length Analysis Figure 1 shows the distribution of the length of text in various languages. Notably, there is a significant variation in the median text lengths among languages. For example, languages such as Somali (som) exhibit a higher median text length compared to others like Algerian Arabic (ary) and IsiZulu (zul), which have relatively shorter median lengths. This disparity suggests that certain languages may inherently require more characters to convey meaning, or it could reflect cultural or linguistic differences in communication styles. Additionally, the spread of the interquartile range (IQR) and the presence of outliers indicate that some languages have more consistent text lengths, while others show greater variability. Languages with wider IQRs and more outliers, such as Nigerian Pidgin (pcm) and Oromo (orm), may be used in more diverse contexts or have more flexible grammatical structures that allow for varied text lengths. This is not surprising for Nigerian Pidgin, as it is a variant of English mixed with several local languages Akande 2010 . Figure 2 extends the