Automatic Extraction of Pronunciation Rules and Generating Pronunciation Variants of Words to Use in Automatic Continuous Speech Recognition