Culture, a pivotal concept in social and behavioral science, elicits diverse interpretations. Insufficient recognition of this variance can hinder constructive dialogue across disciplines. This study aims to understand themes encoded in more than seven hundred definitions of culture collected from across disciplines. Two topic modeling approaches are compared - a novel use of congruence coefficient analysis in the LIWC based Meaning extraction method and Latent Dirichlet Allocation. The former approach was found to be more effective for uncovering themes in definitions of culture as text and resulted in a 16-factor varimax rotated model. In the latter approach a partially coherent 32-factor model was retained. The differences in the methods and results, and the applications of the 16-factor structure as a marker for identifying cultural relevance in texts are discussed.
原文链接
分享
引用
