30 replies to this topic

[BrianR]

This plural vs singular thing with Google - didn't I read somewhere in one of the threads that Google was now stemming and thus relating singular and plural terms, among others??

Or perhaps I was just suffering from a particular nasty case of caffeine deficiency at the time...

BrianR

[bwelford]

This plural vs singular thing with Google - didn't I read somewhere in one of the threads that Google was now stemming and thus relating singular and plural terms, among others??

BrianR, that's what Google would like you to believe, but it isn't true.

Thanks also for the warning re groundhog, but the avatar still seems to have you in mind at this very moment.

[BrianR]

BrianR, that's what Google would like you to believe, but it isn't true.

Ah well, Barry - we can but live in hope...

BrianR

[Grumpus]

What google is doing with the singlular and plural is far more fascinating than just stemming. Stemming assumes that the singular and plural are equal and interchangable. What's happening here is assigning a related but somewhat different meaning to the word (or groups of words).

Here's one I was going to show you all a week or so ago, but it was all messed up with the update. It's a little different now that it's settled out a bit, but the idea is the same.

Here are the terms (in bold) along with the meanings that it appears that Google has assigned to them. (Go ahead and open another window and run the terms in Google to play along at home).

1) I have a whiplash injury
2) I need someone who specializes in whiplash injuries

Though there are many of the same sites in the SERPS for each of these terms, you can easily see a big weight difference. With whiplash injury the weight is put on informational pages - pages that will explain what it is, how you might relieve the pain, where you might find a doctor, etc. When you switch to whiplash injuries the weight transfers over to more compensatory type sites - mainly legal firms and such. True, both still include instances of each other possible definition/desire, and they really contain the same set of documents - they are just sorted a bit differently based upon the "probable meaning" of your search.

The example used earlier in this post does something very similar - it's just harder to see what different meaning Google has put into the term. At first, it looked as if maybe affordable web designer (singlular) was geared toward directories that list various ones and the plural brought specific ones. That doesn't make much sense and doesn't really work when you run the searches, though. I can't figure out exactly what the semantic variance is in that search (though I haven't tried very hard). It's obvious to see that there is one, it's just not as obvious as in the first example. (This term may also be in "heavy play" right now, too - if so, any effort we put into it may be self defeating...)

Anyway, there are plenty of examples of this around. Note that (going back to the first example) the term is treated as a whole. It's not the word "injury/injuries" that has a meaning - it works differently if you put another word with "injury". It's the entire term. It's still early to tell where this is going, but I suspect that if you are running keyterm research, it'll help you to know what Google thinks the term means and optimize for the version of the term that matches Google's definition of it. Then, you'll still show in the "other version" but you'll get a nice boost in your "primary version."

Then again, I could be wrong. As I say, it's early and the whiplash injuries example doesn't work for all terms and most terms don't have an obvious "meaning" yet. We'll see, though.

G.

[bwelford]

That's fascinating, Grumpus.

Of course if you really want to get fancy, you can do a search for
whiplash injury -"whiplash injuries"
and compare that with a search for
whiplash injuries -"whiplash injury"

These two subsets of web pages completely cut out any where stemming might have got the same web page to appear in both lists. I think they support your view on how the singular and the plural versions may be interpreted. What do you think?

[Grumpus]

I hadn't really gotten that deeply into the exploration yet. Yesterday was the first time I'd tried it since Gladys hit where it was starting to work again for that term.

The technique you describe would show what the webmasters of each site decided that the terms meant - which, if we understand how semantics works, is exactly how meanings are derived. Google doesn't actually "know" what terms mean, they know what types of things the pages relate to. In the case of "whiplash injury(ies)", there is a natural tendency to use the variations in certain ways, so the meaning is more easily extrapolated.

"We specialize in helping people with whiplash injuries."
"Here is some information that can help you with your whiplash injury."

In the case of "affordable web designer(s)" the distinction may not be as clear. But, Google has obviously found some difference - we just can't see what it is.

So, yes, Barry. Minusing out the opposing term should be useful in determining what set of pages provides the semantics "seed" for one version - now we just need to figure out what goes into drawing the conclusion that Google is coming to by looking at that set of pages.

Fun fun fun!

G.

[darciusrex]

Grumpus-
I've noticed the same thing about your example in other areas too. Whenever I've searched for the plural of something I've seen sites come up with something to sell and when doing a search for the singular version I've only found informational pages, mostly from manufacturers, with nothing to sell. So is your theory google is trying to train users to search a certain way, if they want info they search for the singular and if they want to shop they search for the plural? It would make sense, since most people comparison shop online and would somewhat instinctively use the plural when wanting to compare, see their options or buy. Fun, fun, fun indeed!

dh

[BrianR]

Grumpus

That is indeed fascinating stuff - I need to keep a watch on that in future searches.

BTW, re the other search query, it occurs to me that G might be stemming on the narrower term 'web designer(s)' rather than on the wider 'affordable web designer(s)' term.

I guess there has to be some sort of limit to the number of words in a term for which the algo can compute stemming - or, perhaps as a non-coder, I'm thinking about this stemming business all wrong...

BrianR

[peter_d]

I've seen that too, Grumpus.

I've been testing for a few months over a wide range of terms and it appears as if Google may be using the invisible buttons that Danny talked about a while ago.

Essentially, the search results will change depending on subtle semantics in the query.

So, phrase matching SEO will only get you so far. The SERPs returned for a search phrase may not feature that phrase at all, rather, they contain terms that are semantically associated with the query by way of a pre-defined keyword set.

Still testing, but it's certainly beginning to look that way.

[BrianR]

Ooh goodie! - if so, this is going to make SEO copywriting:

(a) more difficult = higher fees - hooray!

( more challenging = much more fun - hooray!

BrianR

[Grumpus]

So is your theory google is trying to train users to search a certain way, if they want info they search for the singular and if they want to shop they search for the plural?

Nah. It's not google that decided that plural was commercial and singular was informational. It was the people who built the web sites.

I sell products.
Here is product information.

It's just the way the words are used most commonly. Interestingly, if anything, the "affordable web designer(s)" search seems to be the other way around - it's just how it appears on the most pages.

Good stuff, everyone else. Peter, I'm not sure I'm buying the "hidden buttons" theory, yet. Maybe, but I dunno. I'm not sure about a "keyword set" either. Because everything is different and because it's taking so long for Google to muscle through this rough spot, it feels like they are making up their relational database on the fly rather than starting with a wordlist or anything like that.

Then again, I could be wrong there. Maybe it's a seed list combined with other things? Who knows? It's definitely going to be easier to see once more sectors start to stabilize a bit.

G.

[peter_d]

For those interested in the hidden buttons theory, check out Dannys article here.

http://searchenginew...cle.php/3115131

Stock, the keyword sets are bound to be dynamic. They will be seeded to start with, however, as they have to start with some data.

Edited by peter_d, 10 February 2004 - 08:55 PM.

[Grumpus]

Yup yup. I'll buy the seed list theory all day long.

G.

[Scottie]

Thanks for the link Peter- I missed that one.

[DanThies]

Maybe this is all just the result of Google's click-through tracking. We saw similar types of distinctions on DirectHit back in the day, and in the way this changed MSN and Hotbot's results (which were not pure Inktomi back then), because users on Hotbot (the real data source) were telling them which results were the most relevant.