Automatic Recognition of Spoken Spelled Names for Obtaining Directory Listings

Abstract

An automatic speaker-dependent word recognizer is used to accept strings of spelled letters spoken in isolation. The output of the recognizer is a set of best-candidate letters for each letter spoken in the string. Candidate strings forming spelled last names and initials are compared to name strings obtained from a telephone directory stored in a disk file. A systematic search is carried out to find matching entries in the directory. An evaluation has been carried out with ten talkers spelling lists of names extracted from an 18,000-entry telephone directory. Although the median acoustic error rate per spelled letter is approximately 20 percent, the median error rate in furnishing the requested directory entry is approximately 4 percent.