Synonyms and Search Engines

How smart are they really?

[JunkMail]

Being fairly new to the SEO world, I was pondering a thought:

Can Search Engines such as Altavista, Excite and Google match search keywords with synonyms? Like, if I did a search for puppies, would the S.E. also search for dogs, K-9's, etc. and turn up pages that may not even have 'puppies' in them, but synonyms or other related words?

Thanks for the time and help

[seth_wilde]

As of now, the answer would be no. The closest thing to what you've described would be Infoseeks ability to recognize different versions and tenses of the same word (manufacture, manufacturer, manufacturing, etc.) There has been some rumors of term vector databases being able to recognize synonyms, but I have yet to see any evidence of this. (I think most of these rumors stem from a report talking about automatic directory categorization which I haven't even seen in use yet)

[Brett_Tabke]

Ask Jeeves also does that and Fast is getting ready to put similar online too.

[metaman]

Wouldn't this be considered stemming? Where manufacture, manufacturer, and manufacturing would all be indexed as manufacture (or manufactur?)

[seth_wilde]

It's close, but I believe word stemming is more like partial matching (a search for surf would turn up surfbird, surfboard, surfcasting, etc...)

[engine]

The way the results may be presented from a SE mostly relates to the content in the site. Store the alternatives in your own site (being careful not to inadverently spam the engine).
Linguistic technology has come along really well in recent years, although I'm sure it would raise the bar of complexity significantly.
I've done some research, at the beginning of this year, to try to identify this on the SEs and, as yet, i've found nothing that suggests it works (if it's there at all). I'd guess it is only another layer of complexity for the engineers at the SEs.

My research: With almost 200 searches, studying the top five results in each return from AV, Infoseek/Go and Lycos. That's about 3,000 pages of text with meta tags, the only link was where the meta tags had the information.
I don't fancy doing that lot again.

[Brett_Tabke]

Interesting engine. The big thing to remember is that this is a growing and coming trend. Engines like Ink now have a small team of linguists on staff. Relating kw's and assigning them numbers to generate a vector map of related sites really isn't all that difficult.

[seth_wilde]

"Relating kw's and assigning them numbers to generate a vector map of related sites really isn't all that difficult"

It really is pretty simple and AV has already come up with a way to do it. They spidered the yahoo database and then came up with a common vector for each yahoo category. This then allows them to recognize all common terms for each specific industry.

But from everything that I've read this wasn't developed to be included in the search results. The reason they developed this was so that they could make a directory that automatically sorts sites into the most relevant categories.