Tracing Thematic Shifts in Literature Translation: A Topic Modeling Analysis of Yu Hua’s To Live

Abstract

This study uses Latent Dirichlet Allocation (LDA) topic modeling to compare thematic structures in Yu Hua's novel To Live and its English translation, examining how themes are preserved or altered. After digitizing and preprocessing both texts, separate LDA models yielded six topics for the Chinese original and five for the English version. Key findings show high fidelity for core themes like war and family. For example, war-related topics strongly aligned across versions. However, secondary or culturally specific themes, such as gambling, were diluted or merged in the English translation, while familial themes were consolidated. The research concludes that translation reframes narratives by altering lexical and thematic organization. Topic modeling effectively highlights these shifts, underscoring the translator's role as a narrative architect and offering a valuable method for digital literary and corpus-based translation studies.