The paper introduces the author-topic model, a generative model for documents that extends Latent Dirichlet Allocation (LDA) to include authorship information. Each author is associated with a multinomial distribution over topics, and each topic is associated with a multinomial distribution over words. A document with multiple authors is modeled as a mixture of the distributions associated with the authors. The model is applied to a collection of 1,700 NIPS conference papers and 160,000 CiteSeer abstracts, using Gibbs sampling for inference. The authors compare the performance of the author-topic model with LDA and a simple author model, demonstrating that the author-topic model recovers topics and can be used to compute similarity between authors and entropy of author output. The paper also discusses the generative models for documents using authors and topics, the Gibbs sampling algorithms, and presents experimental results showing the predictive power and illustrative applications of the model.The paper introduces the author-topic model, a generative model for documents that extends Latent Dirichlet Allocation (LDA) to include authorship information. Each author is associated with a multinomial distribution over topics, and each topic is associated with a multinomial distribution over words. A document with multiple authors is modeled as a mixture of the distributions associated with the authors. The model is applied to a collection of 1,700 NIPS conference papers and 160,000 CiteSeer abstracts, using Gibbs sampling for inference. The authors compare the performance of the author-topic model with LDA and a simple author model, demonstrating that the author-topic model recovers topics and can be used to compute similarity between authors and entropy of author output. The paper also discusses the generative models for documents using authors and topics, the Gibbs sampling algorithms, and presents experimental results showing the predictive power and illustrative applications of the model.