Valued search algorith.

[jokke]

Hiya, iwe got a list with 1200 strings and id like a search function where i would be able to search without writing 100% right.

anyone know of anything like this ?

[PsaltyDS]

Hiya, iwe got a list with 1200 strings and id like a search function where i would be able to search without writing 100% right.

anyone know of anything like this ?

By "search without writing 100% right", I assume you mean in-exact matches. To get a computer to do that, you have to define what you mean by "in-exact" so you can implement it in comparison rules. It will get very complicated very quickly if, for instance you want phonetic matching ("phone" ~= "fone").

Google search does something like that if you search for "fone" it will prompt you "Did you mean to search for 'phone'?" But that is very heavy weight programming not to be taken on lightly with a scripting language like AutoIt.

[Toady]

I think your talking about binary search, where not every one of the 1200 string you got are compared. This requires the strings to be sorted alphabetically for it to work. Look at _ArrayBinarySearch() in Array.au3.

[jvanegmond]

Google search does something like that if you search for "fone" it will prompt you "Did you mean to search for 'phone'?" But that is very heavy weight programming not to be taken on lightly with a scripting language like AutoIt.

Actually, that is not very difficult. I once saw on tv how do they do that, they store a list of all search queries and in what order they have been done. Once certain combinations are more common after each other, such as first 'fone' and then 'phone' in correct spelling it takes the word 'fone' up to mean 'phone.

A very straight forward algorithm that can be done in AutoIt, although requires quite a heavy computer to process the amount of queries Google gets.

Edit: On topic, StringRegExp might be abled to do what you want. Use . as a wildcard in the pattern.

Edited July 27, 2007 by Manadar

[PsaltyDS]

Actually, that is not very difficult. I once saw on tv how do they do that, they store a list of all search queries and in what order they have been done. Once certain combinations are more common after each other, such as first 'fone' and then 'phone' in correct spelling it takes the word 'fone' up to mean 'phone.

A very straight forward algorithm that can be done in AutoIt, although requires quite a heavy computer to process the amount of queries Google gets.

Edit: On topic, StringRegExp might be abled to do what you want. Use . as a wildcard in the pattern.

There's the rub though. Millions of people making brazillions of searches give the statistics an input to detect what THE HUMANS did to indicate matches. In your example, it was a statistically significant number of PEOPLE doing something that taught the system the association.

Now tell me how you would implement that associations (and all others) from the start in your code? The OP is talking (unless I misunderstood him, which is very possible) about writing some code now, that will do these searches now, without waiting for millions of people to "teach" it what to look for. That means hard coding near-match logic in the script.

The other problem remains: As sweet as AutoIt is, it can't possibly be the right choice to implement language analysis this complicated.

All this opinionating is moot, of course, if the OP meant something much more modest in scope.

[jvanegmond]

There's the rub though. Millions of people making brazillions of searches give the statistics an input to detect what THE HUMANS did to indicate matches. In your example, it was a statistically significant number of PEOPLE doing something that taught the system the association.

Now tell me how you would implement that associations (and all others) from the start in your code? The OP is talking (unless I misunderstood him, which is very possible) about writing some code now, that will do these searches now, without waiting for millions of people to "teach" it what to look for. That means hard coding near-match logic in the script.

The other problem remains: As sweet as AutoIt is, it can't possibly be the right choice to implement language analysis this complicated.

All this opinionating is moot, of course, if the OP meant something much more modest in scope.

I stand by my point. The algorithm used still is not very complicated.

By the way, over 50% you said, I already stated in my last post. Thanks for repeating.

[PsaltyDS]

I stand by my point. The algorithm used still is not very complicated.

By the way, over 50% you said, I already stated in my last post. Thanks for repeating.

Yes, you win... sigh... I'll admit your sig is much more annoying!

[jvanegmond]

Yes, you win... sigh... I'll admit your sig is much more annoying!

I do my best.