Locating Misspellings, Typos, and Approximations
The approximate pattern matcher lets you find "looks roughly like"
or "sounds like" information. To invoke a fuzzy match precede the
word or pattern with the '%' (percent) character.
This pattern matcher locates items by examining how "closely" the text
matches your query item. It uses percentages of difference as a measure, and
will default to finding items with 80% similar content. You may specify
a different percentage on the query line by following the '%' with a two
digit number that represents the value you desire.
It is important to note our method is far superior to the "soundex" method
found in many other tools. Here's why:
- Handles character transpositions and omissions.
- Can be used on non-word items like addresses.
- Will match foreign language constructions.
- Finds accidental OCR errors or character insertions.
Examples:
Expression Will Find
-------------------------------------------------------------------------
ronald %regan Ronald Raygun, Ronald Re~an, Ronald 8eagan
%75MYPARTNO9045d/6a Anything within 75% of looking like MYPARTNO9045d/6a
mo* %quadaffi mohamar kadaffi, Mo Qadaffi, etc...
Notes:
Since fuzzy searching by its very nature tends to bring back
more than you really wanted, it is always a really good idea to try to
"anchor" your fuzzy term with another keyword. (Like the %regan example).
go back
Copyright © 1996 Thunderstone Software
|