Text this: Use different acoustic model data mining methods, comparing the recognition results on the same database