Theme

This chapter demonstrates how big data and computation can be used to identify and track recurrent themes as the products of external influence. It first considers the limitations of the Google Ngram Viewer as a tool for tracing thematic trends over time before turning to Douglas Biber's Corpus Linguistics: Investigating Language Structure and Use, a primer on various factors complicating word-focused text analysis and the subsequent conclusions one might draw regarding word meanings. It then discusses the results of the author's application of latent Dirichlet allocation (LDA) to a corpus of 3,346 nineteenth-century novels using the open-source MALLET (MAchine Learning for LanguagE Toolkit), a software package for topic modeling. It also explains the different types of analyses performed by the author, including text segmentation, word chunking, and author nationality, gender and time-themes relationship analyses. The thematic data from the LDA model reveal the degree to which author nationality, author gender, and date of publication could be predicted by the thematic signals expressed in the nineteenth-century novels corpus.