Advances in Large Vocabulary Speech Recognition