Unsupervised adaptation device speech recognition using reliable information through the n-best hypothesis.

摘要

Apparatus voice recognition adapts a model of initial voice based on voice input from the user and comprising: a speech model that represents speech in the form of a plurality of models of speech units associated with a plurality of speech units; a device voice recognition processing on voice input from a user using said speech model to generate the N-best recognition solutions; an extraction system reliable information for analyzing said N-best recognition solutions and selecting a candidate set of reliable recognition between said N-best recognition solutions; adaptation system that uses said reliable recognition candidates to adapt said speech model.