Data-Driven Culturomic Analysis of Historical Themes in Fiction Using Decade-Level Topic Modeling

This paper presents a large-scale culturomic analysis of$\mathbf{3 0, 2 5 6}$English-language fiction books published between 1820 and 1929, with the objective of examining whether historical and cultural events leave detectable thematic traces within fictional narratives. Using a decade-level modeling framework, the corpus was processed through a pipeline integrating text normalization, Latent Dirichlet Allocation (LDA), and semantic clustering. The LDA models produced 589 highprobability topic terms, which were subsequently aggregated into five overarching meta-themes using TF-IDF and Non-Negative Matrix Factorization. The results reveal both explicit and implicit cultural signals, including a distinct Napoleonic topic cluster in the 1820s and prominent maritime terminology in the 1840s, corresponding to major geopolitical and exploratory events of those decades. The temporal evolution of the identified meta-themes-Family and Domestic Life, Religion and High Formality, Military and Exploration, National Identity and Institutions, and Youth and Dialogue-further aligns with documented historical and literary trends of the nineteenth and early twentieth centuries. These findings demonstrate that fiction, despite its imaginative nature, encodes measurable cultural patterns and can serve as a viable foundation for culturomic research. The study also highlights the feasibility of using open-source tools to perform large-scale historical text analyses, offering a scalable framework for future work.