Browse Prior Art Database

Phonetic Approximator System for Surname Searching

IP.com Disclosure Number: IPCOM000037328D
Original Publication Date: 1989-Dec-01
Included in the Prior Art Database: 2005-Jan-29
Document File: 4 page(s) / 18K

Publishing Venue

IBM

Related People

Namerow, W: AUTHOR [+2]

Abstract

Directories without a phonetic system require the user to know the several ways a name may be spelled. Since many names are foreign in origin, various spellings for a given sound are possible. This system analyzes and reduces these spellings to two phonetic approximations. This allows the user to type in the name in any spelling which approximates the sound of a surname and still find the name in the directory.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 33% of the total text.

Page 1 of 4

Phonetic Approximator System for Surname Searching

Directories without a phonetic system require the user to know the several ways a name may be spelled. Since many names are foreign in origin, various spellings for a given sound are possible. This system analyzes and reduces these spellings to two phonetic approximations. This allows the user to type in the name in any spelling which approximates the sound of a surname and still find the name in the directory.

The system described provides two levels of phonetic approximation for a given surname. The method was devised to account for various differences in spellings of sounds in English names and in foreign names. The Phonetic Approximator System for Surname Searching (PASS) system accounts for different pronounciations as well. Dialects and regional intonations most often shift the sounds of vowels and dipthongs. In the PASS system these are reduced to single vowels. Consonants which are close in sound, or are shifted in sound in foreign languages, are mapped to the dominant consonant. Semi- vowels (w and h) are removed, but still affect the resulting vowel mix. Various consonant blends are recognized and are reduced to the dominant or most probable consonant sound.

To use the PASS system on a given list of names, the PASS algorithms must be applied to every surname in the list to be searched. It is presumed that the two keys generated by PASS will then be stored in the list. When a search is undertaken, the user will present a surname which is then recoded using the PASS system. The resulting codings are then used as 'test' keys to search the list. Entries in the list that match the first test key are then shown to the user. The user should then find the desired entry in the list or request the other more inclusive search using the second test key.

Any such phonetic approximation or reduction system will yield matches which are more or less distant from the user's desired entry. These are called 'false matches'. This system creates two sets of matches to balance the likelihood of matching the users' desired entry with the number of false matches the user must review.

To do this the PASS system has two levels of recoding. The first is an algorithm which retains much phonetic discrimination and will provide a list of phonetic matches with a reduced probability of false matches. It is called the 'Reduced Probability of False Matches' algorithm. The second algorithm is a complementary recoding which compensates for phonetic matches the first may have missed and also reduces the phonetic discrimination yielding a very high probability of finding the user's desired entry. It is called the 'High Probability of Matches' algorithm. It is intended that the user be presented with any ab- solute spelling matches first, if any exist, then a list developed using the Reduced Probability of False Matches algorithm next, and then High Probability of Match last. This can be done on one presentat...