Path: utzoo!attcan!uunet!decwrl!mips!swrinde!ucsd!rutgers!rochester!pt.cs.cmu.edu!dsl.pitt.edu!pitt!willett!ForthNet
From: ForthNet@willett.pgh.pa.us (ForthNet articles from GEnie)
Newsgroups: comp.lang.forth
Subject: other forth applications
Message-ID: <1744.UUL1.3#5129@willett.pgh.pa.us>
Date: 15 Sep 90 01:22:36 GMT
Organization: String, Scotch tape, and Paperclips.  (in Pgh, PA)
Lines: 48


 Date: 09-13-90 (10:57)              Number: 3769 (Echo)
   To: ZAFAR ESSAK                   Refer#: 3743
 From: GENE LEFAVE                     Read: NO
 Subj: SOUNDEX                       Status: PUBLIC MESSAGE

 ZE>At the moment I am not sure exactly how this function can be useful. 
 ZE>It does group names which at times seems close: 
 ZE>        e.g. SCHMIDT, SMITH, SMYTH are all S530 

 Although I don't pretend to be a SOUNDEX expert I have some experience
 using it.  First, the state of Illinois uses it to generate driver
 license numbers.  A license number is the SOUNDEX code for your last
 name, a first name code, ( I don't know where that comes from), and a
 coded birth date. 

 I used to use SOUNDEX code to retrieve entries in a database.  Using
 SOUNDEX made the program very tolerant of spelling errors.  I seem
 to recall that certain database programs had this function built in.
 However, English has so many short words that I found that in many cases
 I was essentially searching on the first character.   So I went to
 a string search.

 As to the basic algorithm, the idea is to use the first letter, then
 drop all vowels, then group the remaining consonants into 6 sound alike
 classes.  These classes are English specific, not necessarily ethnic.
 adjacent duplicates are dropped. 
        SCHMIDT  =  S530      because
              S    first character.
              C    dropped because its same class as S and adjacent.
              H    always dropped
              M    class  5
              I    dropped vowel
              D    class 3
              T    dropped, adjacent class 3

 You can easily work out the other names.   Its useful for names because
 most last names are long enough to generate a meaningful code.  Assuming
 a list of 1,000,000 names SOUNDEX hashes to 5616 codes, for 180 average
 collisions, which would not be difficult to resolve with a first name
 and birthdate, or some other type of qualifier.  You have to remember 
 that it was originally set up for manual searching.

 ---
  ~ EZ-Reader 1.13 ~ 
-----
This message came from GEnie via willett through a semi-automated process.
Report problems to: uunet!willett!dwp or dwp@willett.pgh.pa.us