This download contains a portion of the data for the "Dialog State Tracking Challenge" (DSTC). Specifically, it contains data and labels for the "train3" set. The source data is based on the "Spoken Dialog Challenge" corpus, published by Carnegie Mellon University, with labels and some transcriptions added by Microsoft. For more information about the DSTC, including pointers to the other datasets, see http://research.microsoft.com/en-us/events/dstc/