Providing users with 'did you mean to type' spellcheck ....

Question

I had a request come in to allow the end user to type in a value - and my program would search a particular field and tell the user if that value existed or did not exist. fine. they also want to know if there's anyting "similar". so, sure, i can do: field = user_input_variable or field like %user_input_variable% - but is there any mechanism already out there - or what about a pl/sql package someone wrote and is sharing (?) that would catch records in a spellcheck fasion, where ... maybe if 3 or more consecutive characters matched - or two digits were transposed? they're not typing in words out of a dictionary, they're typing in values that are specific to us.
thanks in advance!

Answer 1

Ron Mey-Ami wrote:
I agrere with Justin, the UTL_Match package sounds like the best choice (Though I personally haven't used it but rather used SOUNDEX since I am working with a prior oracle version)Just to clarify, you are comparing apples and oranges. SOUNDEX and UTL_MATCH are based on completely different methods and serve different purposes. Soundex lets compare words that are spelled differently, but sound alike in a specific language+*. UTL_MATCH.EDIT_DISTANCE simply gives you number of characters that are different between two strings. Therefore, SOUNDEX is NLS dependent while UTL_MATCH is not. SOUNDEX that comes with Oracle is for English language only. Based on nick ukraine.believe, OP would have to find Ukrainian SOUNDEX for Oracle (assuming such is available). Another major difference is UTL_MATCH.EDIT_DISTANCE provides a way to measure how different words are while SOUNDEX simply tells if words sound alike - it does not provide any way to measure how different words sound. Anyway, below is an example:
SQL> SELECT utl_match.edit_distance('expresso', 'espresso') DIST,
  2         SOUNDEX('expresso'),
  3         SOUNDEX('espresso')
  4    FROM dual
  5  /
      DIST SOUN SOUN
         1 E216 E216
SQL> SELECT utl_match.edit_distance('expresso', 'e1presso') DIST,
  2         SOUNDEX('expresso'),
  3         SOUNDEX('e1presso')
  4    FROM dual
  5  /
      DIST SOUN SOUN
         1 E216 E162
SQL> As you can see, UTL_MATCH gave us same smallest possible (except exact match) distance in both cases since it olny takes one change to transform first word into second word. SOUNDEX results in second case show words do not sound alike. Below is an opposite example:
SQL> SELECT utl_match.edit_distance('expresso', 'eakspreasaah') DIST,
  2         SOUNDEX('expresso'),
  3         SOUNDEX('eakspreasaah')
  4    FROM dual
  5  /
      DIST SOUN SOUN
         7 E216 E216
SQL> SY.