Well my site is open for public for a couple of days now and I'm getting a nice stream of visitors through some advertisement. I have installed the Search Contribution and I would just like to recommend it to everyone: it is absolutely vital!

Looking through the first 30 search terms, the number of spelling mistakes is absolutely amazing... no wonder some people never find what they want, you are not likely to get the right results if you search for "Collin Mcray" if you mean "Colin McRae"! Thanks to the Search Contribution you can catch these foolish typing errors and present the visitor with "Did you mean... Colin McRae?" instead of "nothing found, you dumbarse".

Thought I'd share this with you

p.s. check your keywords several times a day! Correct typing errors or alternative methods of writing the same word.

Link to this contrib. please?

I have to agree that this is a most excellent contribution. It's pretty easy to install and works pretty well!

Here it is Search Enhancements

This is just really bizarre... with over 70 keywords about 80% are typed wrong... do you also get the impression sometimes that the entire planet is populated by fools hehehe

You could always try using the soundex function of mysql

http://www.mysql.com/doc/en/String_functions.html

SELECT SOUNDEX('Hello');
-> 'H400'

so in a product search you could have something like

SELECT pd.products_name
FROM products_description pd
WHERE
SOUNDEX(pd.products_name) = SOUNDEX($UserEnteredCriteria);

----
SELECT SOUNDEX('Colin McRae');
->C4526

SELECT SOUNDEX('Collin Mcray');
->C4526

so in terms of soundex both of these "sound" the same and would be found regardless of spelling as long as it sounds correct

oh and don't be too harsh on your customers, I bet you guys never make any spelling mistakes when searching for stuff on google

I had no idea this was possible... in fact I have no idea how the soundex thing works, surely the computer doesn't know what stuff sounds... nor can it access a huge database where everything is stored. I don't know how it works, but it sounds very interesting indeed.

So much stuff I haven't heard of before

Regards,
Willem.

p.s. to answer your question... no, I don't make mistakes in the products I'm looking for honestly. Perhaps you might type "fifa 2004" instead of "Fifa Football 2004" but that's hardly a typo...
When people start typing 'games' on a gameshop website, then I'm starting to get worried actually hehehe

Good tip. A pitty for me that this function is optimized for the English language only. I have created a php function like soundex for Dutch, very simple, using lots of string replacements. I haven't tried to implement it in osC though. I am not a good programmer, so it is all very time consuming.

To create a language independand (spelling ok? ), you could use the PHP levenshtein (or was it lhevenstein or levensthein? I always forget) function. It is not as advanced as soundex, but I think it would do quite good job for every language. http://nl.php.net/manual/en/function.levenshtein.php

But don't ask me how! I only know should not be to difficult for a smart programmer.

This post has been edited by paulm2003: Nov 15 2003, 01:23 PM

The lebensthein thing I can't even understand from reading the manual

Anyway, the search engine could use a lot of work, besides the search enhancement is merely a start, not a finished job. With my customised website I have made it possible to actually correct several typos in one searchstring and take the user directly to the results he wants (so Collin Mcray gets you to search results with Colin McRae Rally 4, and so does colin mcrea). This is based on common sense (search replacements), not soundex. Unfortunately it's not the neatest coding and highly focussed on my custom website... if I get some time I might try my hands at a contribution.

This post has been edited by WillemB: Nov 15 2003, 07:14 PM

Soundex codes are based upon sounds. Names having similar pronunciations are ideally assigned the same code, a series of four characters. The first character is the initial letter used in the search, in the case of Colin McRae that would be a C. The last three characters are numbers based upon the other letters of the search after removing vowels. The three numbers are coded according to the following relation:


1= B, P, F, V
2= C, S, K, G, J, Q, X, Z
3= D, T
4= L
5= M, N
6= R

H,Y and W are ignored and double letters are treated as a single.

So if you take Colin McRae this turns into CLMCR, then changing the last characters you get C4526. Generally soundex codes should only be 1 letter and 3 numbers so you should really truncate that returned by mysql to be C452. But anyway that is how it works, no large lists of codes stored etc.

And to be honest I don't always know how to spell Colin McRae and I'm Scottish, it could easily have 2 l's or McRae could be spelt MacRae. Using soundex it doesn't matter because they all "sound" the same.

This post has been edited by mikeq: Nov 16 2003, 01:34 PM

From what I understand metaphone functions are similar to soundex but more reliable. Soundex was created to prevent misspellings in Names for genealogical records where metaphone takes an entire language into consideration.

Both PHP and MySQL support Metaphone as well as Soundex functions.

But remember: the results are not the same for PHP and MySQL (not sure if that's true for both soundex and metaphone but at least for one of those).

If you want to try a metaphone alike function live, take a look at http://www.eeweb.nl/memory/pn/
I never really got to finish the site but it does work. The English version is best, but also Dutch and Norwegian (a little) is supported

PaulM

P.S.
I made this function for a total different purpuse, it has to do with remembering numbers. If know how to translate numbers to words (actually to sounds and with the sounds you create words), you will be able to remember numbers much better.

So which file and where in that file would I add the soundex function?

Thanks in advance.