Learning Structural SVMs with Latent Variables

We present a large-margin formulation and algorithm for structured output prediction that allows the use of latent variables. The paper identifies a particular formulation that covers a large range of application problems, while showing that the resulting optimization problem can generally be addressed using Concave-Convex Programming. The generality and performance of the approach is demonstrated on a motif-finding application, noun-phrase coreference resolution, and optimizing precision at k in information retrieval.