Beefy Boxes and Bandwidth Generously Provided by pair Networks
Perl: the Markov chain saw

Fuzzy searching

by jaxz (Initiate)
on May 18, 2000 at 19:41 UTC ( #13032=perlquestion: print w/replies, xml ) Need Help??
jaxz has asked for the wisdom of the Perl Monks concerning the following question:

I have a flat datafile containing a variety of fields, mostly with text data to which I have written a CGI access program that appends the file with user inputted data. Before the file is updated, I want to run a check to see if the data being entered has been repeated elsewhere. The trick is that although the data might not have been repeated in the exact same form, I want the check to see if the newly entered data looks *similar* to existing data rather than exactly the same. So I'm looking for a fuzzy search algorithm, nothing too complex, but robust enough to catch possible repetitions in the data. Any help, greatly appreciated... jaxz

Replies are listed 'Best First'.
Re: Fuzzy searching
by lhoward (Vicar) on May 18, 2000 at 20:16 UTC
    turnstep's suggestion is excelent. There are 2 other modules you may want to check out:
    1. Text::Soundex - good for determining is words sound-alike
    2. Text::Metaphone - better algorithm that Text::Soundex and good for phrases
    You didn't say in your question if the "similar data" you're trying to match is strings or not. If it is numeric a simple +- %5 type algorithm would work well for determining similarity. For comparing strings one of the simplest algorithms you can use is ignoring case ("ANOTHER MAN DOWN" and "Antoher Man Down" are considered the same thing). Its not very powerful but it is very easy to do.

    Different algorithms may work better depending on exactly what piece(s) of data you want to de-duplicate on. If you're deduplicating on multiple fields some sort of hybrid de-duplication algorithm may be best. Here's an example deduplication scheme I cooked up for a database of people, where they live, and what their income is:

    (String::Approx of LAST_NAME the same) and (INCOME within %5) and (STATE the same)
    In my experience coming up with a de-duplication scheme for user-entered data is easy. Coming up with a good one is hard and may take weeks or months of tuning.
      Thanks for the hints - it sounds like the metaphone module might be what I need as the data is text, and in sometimes quite long phrases where the variation might be word order, similar but not necessarily all matching words and so on. I'll look it up. The ignoring case would not really be sufficient in this instance.
Re: Fuzzy searching
by turnstep (Parson) on May 18, 2000 at 19:53 UTC

Log In?

What's my password?
Create A New User
Node Status?
node history
Node Type: perlquestion [id://13032]
Approved by root
and all is quiet...

How do I use this? | Other CB clients
Other Users?
Others chilling in the Monastery: (6)
As of 2017-07-23 03:20 GMT
Find Nodes?
    Voting Booth?
    I came, I saw, I ...

    Results (343 votes). Check out past polls.