Hierarchical Bayesian models for applications in information retrieval

We present a simple hierarchical Bayesian approach to the modeling collections of texts and other large-scale data collections. For text collections, we posit that a document is generated by choosing a random set of multinomial probabilities for a set of possible "topics," and then repeatedly generating words by sampling from the topic mixture. This model is intractable for exact Probabilistic inference, but approximate posterior probabilities and marginal likelihoods can be obtained via fast variational methods. We also present extensions to coupled models for joint text/image data and multiresolution models for topic hierarchies.