Text this: Hierarchical connectionist acoustic modeling for domain-adaptive large vocabulary speech recognition /