Prev: Generating ordered subsets of fixed size. Was Re: Algorithms?
Next: Ariadne Designs Ltd
From: Anthony White on 15 Jul 2010 16:16
Hello all,

I have a problem at work. I'm a software developer at a small company
(I'm in fact the only developer). I'm basically trying to automate a
lot of the tedious work that the other teams have to do.

What I've been asked is to implement some kind of fuzzy logic input
rules. Basically, we have search lists that have to be searched on
various databases. What we want to do is generate these search lists
automatically based on common misspellings, variations of characters
(such as changing \ to /), and other simple substitutions (& to and,
for example). Can anyone suggest a solution for this? Does anything
exist that can do this as a separate application, and export a list of
terms?

Apparently, the consultant that was spoken to before I was hired told
us that applications or libraries like this exist. After some
searching, I've found nothing.

Any help is appreciated.
From: Ben Bacarisse on 15 Jul 2010 17:40
Anthony White <anthony.white.j(a)gmail.com> writes:
<snip>
> What I've been asked is to implement some kind of fuzzy logic input
> rules. Basically, we have search lists that have to be searched on
> various databases. What we want to do is generate these search lists
> automatically based on common misspellings, variations of characters
> (such as changing \ to /), and other simple substitutions (& to and,
> for example). Can anyone suggest a solution for this? Does anything
> exist that can do this as a separate application, and export a list of
> terms?
>
> Apparently, the consultant that was spoken to before I was hired told
> us that applications or libraries like this exist. After some
> searching, I've found nothing.

This is how some spell-check correction systems work. You could look
at the code for, say, GNU aspell.

Another line of attack (if your database supports it) is to change the
input into a regex:

up/down and left/right

becomes

up[\/]down (and|&) left[\/]ri(ght|te)

I don't know of any software to do this for you though. The regex might
turn out to be faster than searching for all the alternatives one after
the other since every simple alternative doubles the number of match
strings.

These are just random thoughts. Text search experts will probably have
this sort of thing off pat.

--
Ben.
From: Walter Banks on 17 Jul 2010 09:37


Anthony White wrote:
>
> Hello all,
>
> I have a problem at work. I'm a software developer at a small company
> (I'm in fact the only developer). I'm basically trying to automate a
> lot of the tedious work that the other teams have to do.
>
> What I've been asked is to implement some kind of fuzzy logic input
> rules. Basically, we have search lists that have to be searched on
> various databases. What we want to do is generate these search lists
> automatically based on common misspellings, variations of characters
> (such as changing \ to /), and other simple substitutions (& to and,
> for example). Can anyone suggest a solution for this? Does anything
> exist that can do this as a separate application, and export a list of
> terms?
>
> Apparently, the consultant that was spoken to before I was hired told
> us that applications or libraries like this exist. After some
> searching, I've found nothing.
>
> Any help is appreciated.

For fuzzy logic data base work the single best starting point
is any of several books by Earl Cox. Well written and many
good code examples.

My company makes some fuzzy logic implementation tools that may
or my not be useful in the actual implementation.


http://bytecraft.com/Fuzzy_Logic

Regards,

--
Walter Banks
Byte Craft Limited
http://www.bytecraft.com

--- news://freenews.netfront.net/ - complaints: news(a)netfront.net ---
 | 
Pages: 1
Prev: Generating ordered subsets of fixed size. Was Re: Algorithms?
Next: Ariadne Designs Ltd