Gossamer Forum
Home : Products : Gossamer Links : Version 1.x :

Stop Words for Searches

Quote Reply
Stop Words for Searches
I'm not exactly sure how the Stop Words work in Search.PM. It seems that they cause a problem when searching with the boolean AND.

In other words if "in" is designated as a stop word, and then a search is done in the AND mode with something like "Hotel in California", then it won't return a link with the description that has both "Hotel" and "California" in it because the term "in" has been taken out for indexing. Is that right?

I was hoping that the system worked in such a way that a search with AND with a search like "Hotel in California" would return everything that had both "Hotel" and "California", and that it would just ignore the common word "in". Am I missing something here, or does anyone know a way to do this?

--jwm

Quote Reply
Re: Stop Words for Searches In reply to
That's the way it should work.

Searches are tokenized, then scored in first-order ranking.

That means, if you enter "Hotel in California" if a link has

"Hotels that might or might not be great for California visitors"

would show up ahead of a link that had:

"California has great hotels!"

If it's not working that way, then it's a bug, but this is how it was supposed to work, when the 'score' order was set up. It's an in-order search, rather than a true 'phrase' search.

Words are not indexed by phrase, so searching by phrase requires sophisticated algorithms and indexes, or it requires full-text searching.

Make sure you've weighted and re-indexed all the fields you think you are searching on.

http://www.postcards.com
FAQ: http://www.postcards.com/FAQ/LinkSQL/