Probabilistic topic models, built on methods in machine learning and natural language processing, are used to “discover the hidden thematic structure in large archives of documents” (Blei, 2011). This seminar will introduce two probabilistic topic models: latent Dirichlet allocations (LDA; Blei, Ng, & Jordan, 2003) and the correlated topic model (Blei & Lafferty, 2007). Also, an illustrative example of applying these models will be given. All topic modeling analyses will be done in R (R Development Core Team, 2013) via the tm (Feinerer & Hornik, 2014) and topicmodels (Grün & Hornik, 2013) packages. The tm package is used for pre-processing (e.g., reducing similar words to their common root and removing item numbers and punctuations) texts
