Bayesian update of dialogue state (BUDS) is a state-of-the art system for human-computer conversation in dialogues. Here, it is employed to build a speech-driven intelligent assistant. The system manages the conversation to help the user achieve their goal as quickly as possible. The main challenge is to converse in a way that overcomes mistakes made by the speech recognizer, or ambiguous utterances by the user. The system can ask for confirmations, pose choices, and ask for additional information, all in order to gain certainty while maximizing dialogue utility.

The system contains a long machine learning pipeline. It preserves a large number of speech recognition hypotheses by representing them as a confusion network (a compact form of an HMM lattice), and applies a semantic decoder directly to this network. The dialogue state is tracked via a Dynamic Belief Network. The system chooses actions according to a policy that has been learned using a POMDP. The ability of the system to maintain uncertainty significantly improves dialogue utility compared to rule-based dialogue systems.