Another Semantic Tweak

Over the last few days...

[caveman]

A look at our log files from the past few days plus a bit of analysis suggests that G just went a bit more literal with its word forms. Plurals and other forms that we've been capturing for a while now at various sites (where the plurals are not really the focus) just dropped about 5 to 10 spots in a whole bunch of SERP's, especially in the more competitive areas.

I decided to post since it's the sort of tweak that can effect many sites/pages in fairly literal ways. FWIW. Anyone else seeing it?

[nuevojefe]

Anyone else?

[steveb]

I have a term that can be spelled two different ways, kinda like "color" and "colour", which I just noticed was ~ed so that color is now linked to colour, but strangely colour is not associated with color. The rankings have also changed for both terms, but I haven't figured out what the impact should be, other than there would be one.

[RFranzen]

Taking off from steveb's point, Google clearly associates words like color:colour and grey:gray, as well as singular and plural forms. The SERPs using such terms are not exact, though. Rankings tend to shift depending on the spelling used.

Wow, did a comparison with a singular and plural form, and i confirm caveman's observation. They are still associated (even if, say, the plural form is not actually used on a page), but ranking is a lot different now between the two searches -- like hundreds of places.

-- Rich

[RFranzen]

A related semantic point is that "ignored words" are not fully ignored. E.g., the results vary if searcher types:

* <adjective> <noun>
* the <adjective> <noun>

If "the" or "a" is included, Google will tell you it is a common word, and it was not included in your search. Perhaps not, but it does alter the order of results.

-- Rich

[leppjc]

I watch one keyword & the plural very closely and the SERP results have been fairly consistent, in a bad way, for the last 6 months. But have gotten slightly worse this week.

Singular Widget: #1 for my website (for over 10 months) and #3 to #5 for major competitor.

Plural Widgets: last week in the 30's for my website and #4 to #6 for same competitor.
Widgets: this week in the high 40's for my website and still #4 to #5 for same competitor.

I've seen this kind of flucuation for the past 6 months at least. Doesn't seem to be anything new for our results.

[Skier]

A new client caused me to check the SERPs for a category I have not looked at before. This may be old news, but a surprise to me yesterday:

Search "location jewelry"
Results #3-#5 - Lord of the Rings (showing at the location theater)

I tried this for other locations and it seems to happen only when there are no/few jewel stores in the location. Results for larger centers look normal.

If G can get from "jewelry" query to "Rings" results, G's semantic abilities are further developed than I had realised.

[caveman]

Our observation is that at least two sets of sematically-related activities have been occuring since Florida:
1) Evolution of how kw's are related, probably do to more information over time;
2) Tweaks to the algo that are not related to gaining more infomation.

This seemed to be a tweak.

Agree with steveb that it is not limited to plurals; also includes other close-in word forms.

WRT plurals, if searchs on "widgets" and "widget" previously produced very similar results, starting last week, the results are in some cases now are quite different. The new SERP's, in this one resect, seem more similar to Y!, suggesting the possibility that G got more literal to on-page factors. That's not all bad, but would seem on the surface to be a step back from their previously more interpretive results.

Also, not everything we look at with this tweak makes sense to us yet. But then again, that's true with nearly everything G does. It's probably just us. We don't have a single PhD working here in the cave. :-)

[nuevojefe]

I just noticed was ~ed so that color is now linked to colour, but strangely colour is not associated with color.

yea, i've noticed that before, it can often be a one way thing. Interesting though...

[Nikke]

The plural seems to be gone. Searches for seasonal widget and seasonal widgets are showing me very different results.

[walkman]

Ok, I need help. Do the linked (via semantics) words count as one, when it comes to density?
Example(I searched g for ~hotel):
If I have a page, with
City Hotel
City Hotels
City INN

is the "City Hotels" density X or 3X?

[Rugles]

I think my backlinks have changed this morning. Anyone else?

[shri]

caveman, there is a difference in a couple of areas where I track, between "widget reviews" and "widget review".

[starlygirl]

If I search for "mykeyword clothing", the results show "mykeyword", "clothing" AND "clothes" in bold.

If I search for "mykeyword clothes", the results show "mykeyword" and "clothes" in bold, but not "clothing".

The SERPs are different with the two searches as well. Not sure what all that means - I'll leave it to y'all to analyze :-)

[edited by: starlygirl at 5:51 pm (utc) on Oct. 28, 2004]

[jetboy_70]

A site that only mentioned 'widget details' on-page, but always ranked first page for 'widget details' and 'detail widgets'. At the time the toolbar PR updated the ranking for 'detail widgets' disappeared but 'widget details' moved up a couple of places. This returned to how it was originally last week.

I'd assumed this was an algo tweak relating to the order of keywords, but it could easily be explained by a change in the handling of plurals.

[caveman]

>there is a difference in a couple of areas where I track, between "widget reviews" and "widget review"

Meaning, a difference now versus a few weeks ago?