This chapter demonstrates how feature-based models can be extended and used for query expansion using a technique known as latent concept expansion (LCE). The approach has three key benefits, including the ability to go beyond the bag of words assumption, the ability to employ arbitrary features during the query expansion process, and the ability to expand with a variety of concept types beyond unigrams. In addition to the basic LCE model, the chapter also describes a number of powerful extensions, including generalized LCE and LCE using hierarchical MRFs that encode document structure during the expansion process.