The Author-Topic Model and the author prediction

The author-topic model is a generative model for documents that extends Latent Dirichilet Allocation to include authorship information, which is proposed by Michal Rosen-Zvi et al. The model connects each author to a multinomial distribution over topics and associated each topic with a words’ multinomial distribution. A document with multiple authors is modeled as a distribution over topics that are a mixture of the distributions associated with the authors. In this project, I re-implement the model to a collection of about 250 NIPS conference papers (be chosen randomly from a collection of about 1700 NIPS papers). Exact inference is intractable for these datasets and I use Gibbs sampling to estimate the topic and author distributions. The tagging results with different topic numbers are given. After getting the distribution values, I present a new method that apply maximum likelihood estimate to do author prediction on about other 100 papers of which the authors are in the same set as the training papers. The precision of prediction is given. Key word: author-topic model; Gibbs sampling; multinomial distribution; tagging; author prediction;