The new “Language Understanding Intelligent Service” from Microsoft is a pure “do-engine”, and not a dialog system at all. It interprets, or extracts intents, from natural language, such as voice commands, which can then be keyed to actions, for the Internet of Things, wearable devices, etc.
According toof Microsoft Research:
LUIS lets you build an LU model specific to your domain. That LU models takes text (the “question”) as input and produces JSON as output. The JSON contains 2 things: the “intent” of the whole question, and 0 or more entities, which are substrings. The JSON does not contain the “answer” to read back—rather it is classifying the input into categories that you specify. Your app can then use the “category” of the input (rather than the raw input text) to decide what to do.
See also the following forum thread at Chatbots.org: