LANGUANGE MODEL CLUSTERING BASED SPEECH RECOGNITION APPARATUS AND METHOD

摘要

A language model clustering-based speech recognition apparatus and method thereof are disclosed. According to an aspect of the present invention, there is provided a speech recognition method based on language model clustering, comprising the steps of: outputting a first best candidate list by performing speech recognition using input speech data using a first language model; Determining a mixer weight value for each language model community using one of the recognition candidates of the first best candidate list; Interpolating each language model cluster based on the mixer weight value; And recalling the first best candidate list using the result of interpolating each language model ensemble.