The AuthorTopic Model
The authortopic model is a generative model for authors and documents
that reduces the generation of documents to a simple series of probabilistic
steps. Each author is associated with a mixture over topics, where topics
are multinomial distributions over words. The
words in a collaborative paper are assumed to be the result of a mixture of
the authors' topics mixtures.
The results presented on this webpage are extracted from a
single MCMC sample, for a 300topic model for CiteSeer, a 200topic model for Enron
emails, and a 100topic model for NIPS papers.
Information about the Data Sets
AuthorTopic Modeling Results
Applications of the AuthorTopic Model
References

Finding scientific topics, T. Griffiths and M. Steyvers, Proceedings of the National Academy of Sciences, 2004

The authortopic model for authors and documents,
M. RosenZvi, T. Griffiths, M. Steyvers, P. Smyth,
Proceedings of the 20th Annual Conference on Uncertainty in Artificial
Intelligence, 2004.

Probabilistic
authortopic models for information discovery
M. Steyvers, P. Smyth, M. RosenZvi, T. Griffiths,
Proceedings of the Tenth ACM SIGKDD Conference, 2004.