Dialect-based Speaker Classification of Chinese Using Structural Representation of Pronunciation