Using a Pronunciation Dictionary and Phonetic Rules for Name Matching Applications

Phonetic matching is used to identify strings that are likely to have similar pronunciation, regardless of their spelling. The well-known phonetic matching techniques such as Soundex, although simple to compute, are in practice less effective than nonphonetic measures such as edit distances. In this paper we analyse the problem of phonetic matching to explain why Soundex-like techniques are unlikely to work well, and propose a more sophisticated technique for phonetic matching based on information derived from a pronunciation dictionary. This technique is not yet efficient enough for a production system, but is more effective than Soundex and Phonix.