Title: Introduction to Topic Models Automatic segmentation and identification of topics in natural language text and speech has been of much recent interest. Many approaches have been suggested for analyzing the topicality of a text corpus for a number different target tasks, including similarity, classification, summarization, and others. Topic models generally view text as a sequence of observations explained by a sequence of latent topic assignments. In generative topic models, text is produced by mixing together topics selected automatically by a maximum-likelihood training procedure. I will review some classic approaches for topic analysis and thus motivate an in-depth discussion of the evolution of modern topic models over the past decade. I will also discuss topic segmentation models, which break a stream of text or speech into topic-coherent segments without necessarily providing topic identities.