Exploring personality structure in South Africa: A text mining approach

Personality traits provide insight into physical expression, behaviour, and social relations, making personality research vital for both clinical understanding and social analysis. Much of trait psychology developed through the initial analysis of the English lexicon to identify dominant personality traits which then informed the creation of personality assessments like the NEO Personality Inventory-Revised-R. Following this, cross-cultural psychologists used bottom-up approaches in other cultures to identify dominant personality traits and developed instruments like the Cross Cultural Personality Assessment Inventory-2 and South African Personality Inventory. Recently, naturally occurring data like social media posts have been used to study personality. However, much of the analysis for these studies was manually driven using qualitative or quantitative techniques. This study applied text mining techniques, including parts-of-speech tagging and supervised and unsupervised topic modelling, to 60 South African literary texts to identify personality trait terms. While unsupervised topic modelling showed limitations, guided thematic clustering through supervised LDA topic modelling aligned well with the empirically accepted five factors, with mismatches reflecting known issues in fitting certain constructs to African contexts. Results also revealed gender differences in personality consistent with existing literature. Despite the relatively small corpus, the findings suggest that text mining holds promise as an analytic approach for exploring personality structure, particularly for analysing large, naturally occurring datasets in culturally contextual ways.