Topic-Partitioned Multinetwork Embeddings

Citation:
Krafft, P., J. Moore, H. Wallach, and B. Desmarais, "Topic-Partitioned Multinetwork Embeddings", Advances in Neural Information Processing Systems Twenty-Five, Lake Tahoe, NV, 2012.

Abstract:

We introduce a joint model of network content and context designed for exploratory analysis of email networks via visualization of topic-specific communication patterns. Our model is an admixture model for text and network attributes that uses multinomial distributions over words as admixture components for explaining email text and latent Euclidean positions of actors as admixture components for explaining email recipients. This model allows us to infer topics of communication, a partition of the overall network into topic-specific subnetworks, and two-dimensional visualizations of those subnetworks. We validate the appropriateness of our model by achieving state-of-the-art performance on a prediction task and semantic coherence comparable to that of latent Dirichlet allocation. We demonstrate the capability of our model for descriptive, explanatory, and exploratory analysis by investigating the inferred topic-specific communication patterns of a new email data set, the New Hanover County email corpus.