In this article, topic modeling is used to explore a corpus of political speeches by Xi Jinping. It is shown how to preprocess data and how to analyze it using the latent Dirichlet allocation algorithm described by Blei, Ng, & Jordan (2003). Furthermore, it is discussed how the statistical measures of perplexity and coherence can help the researcher to define the number of topics to be analyzed by the LDA algorithm. The results show that two topic models, one with four topics and one with five topics, fit the data best.