Strategies for name recognition in automatic directory assistance systems

The commercial viability of automating large scale directory assistance is shown by presenting new results on the recognition of large numbers of diff erent names. Satisfactory recognition performance is achieved by employing a stochastic combination of N-best lists retrieved from multiple user utter ances with the telephone database as an additional knowledge source. The st rategy is used in a prototype of a fully automated directory information sy stem which is designed to cover a whole country: After the city has been se lected, the user is asked for first and last name of the desired person and , if necessary, also for the street or a spelling of the last name. Confide nce measures are used for an optimal dialogue flow. We present results of d ifferent recognition strategies for databases of various sizes with up to 1 .3 million entries (city of Berlin). The experiments show that for cooperat ive users more than 90% of all simple requests can be automated. Despite th e fact that in the field a lot of practical problems like database or lexic on management or acquainting users with the new systems have to be overcome , the authors nevertheless deem the technology to be highly relevant for co mmercial deployment. (C) 2000 Elsevier Science B.V. All rights reserved.