Database preprocessing and human-interface issues in reverse directory assistance (ACNA) services

Speech synthesis now profitably automates services by speaking information from computer databases. One particularly successful telecommunication service, automated customer name and address (ACNA), requires synthesis with high intelligibility and name pronunciation accuracy, both of which are achieved by current synthesis technology. However, even the best of current technology is not good enough to mindlessly "drop" into complex services. Customized directory preprocessing is necessary to transform listing data, which commonly contains unconventional abbreviations, unlabeled acronyms, and scrambled word ordering, into a sentence suitable for synthesis. We describe our directory preprocessing programs used in successful implementations of synthesis in two major US telephone companies. We also describe an experiment that determined whether the naturalness of recorded speech for prompts and other fixed messages offsets the undesirable discontinuity between natural and synthesized utterances.