LANGUANGE MODEL CLUSTERING BASED SPEECH RECOGNITION APPARATUS AND METHOD

摘要

Disclosed are a device and method for recognizing a sound based on a language model clustering. The method for recognizing a sound based on a language model clustering according to an aspect of the present invention comprises the following steps: outputting a first N-best candidate list by performing a sound recognition by using a first language model for inputted sound data; determining a mixture weight value for each language model cluster by using any one recognition candidate from the first N-best candidate list; interpolating each language model cluster based on the mixture weight value; and rescoring the first N-best candidate list by using a result in which each language model cluster is interpolated.