An Algorithm for Variable-Length Proper-Name Compression

Abstract

Viable on-line search systems require reasonable capabilities to automatically detect (and hopefully correct) variations between request format and stored format. An important requirement is the solution of the problem of matching proper names, not only because both input specifications and storage specifications are subject to error, but also because various transliteration schemes exist and can provide variant proper name forms in the same data base. This paper reviews several proper name matching schemes and provides an updated version of these schemes which tests out nicely on the proper name equivalence classes of a suburban telephone book. An appendix lists the corpus of names used for algorithm test.