Forum Moderators: open

Message Too Old, No Replies

Singular and plural keywords, Google Stemming

         

sandalwood

8:53 am on Nov 27, 2003 (gmt 0)

10+ Year Member



I don't know if this dance is over, but i am sure G has changed so many things in it's algo.

I have just noticed something:

I did a search with kw(in singular) and inside the first page results i noticed that there were titles with kw(in plural) marked bold.
keyword: single holidays results: single + singles holidays
Has G the ability of understanding the singular and the plural of the kws?

It's the first time i notice that in G.
Has anybody noticed something similar?

nileshkurhade

6:36 pm on Nov 27, 2003 (gmt 0)

10+ Year Member



Want to see more irrelevant results full of spam try "transcriptions software" and check out the second page. Its full of spam.

rfgdxm1

7:38 pm on Nov 27, 2003 (gmt 0)

WebmasterWorld Senior Member 10+ Year Member



>Perhaps that is a part of the reason for the last crazy update?

This may explain many cases. For example, if hypothetically your page the was #1 for "dogs" before, it may now be surpassed by other pages that were only well optimized for "dog".

erwig

10:22 pm on Nov 27, 2003 (gmt 0)

10+ Year Member



I have noticed that Google now returns pages that don't even contain the search term. Do a search on Google for 'spanish translation' and look at number 7, El Sagrado Corán. It's a Spanish translation of the Koran, that contains no English. However, when you check the backlinks, many pages link to it with 'Spanish translation of Koran' (or similar) in the link. It seems like Google is putting even more weight on links now.

rfgdxm1

10:32 pm on Nov 27, 2003 (gmt 0)

WebmasterWorld Senior Member 10+ Year Member



Google has always done that erwig.

LateNight

5:30 am on Nov 28, 2003 (gmt 0)

10+ Year Member



Is it possible these massive drops are because the filter has been tripped because it now combines the plural and singular version of widget into one word - that used to be two? If your page has a liberal sprinking of singular and plural versions of a keyword the algorithm now penalizes for repitition and increased kw density?

[edited by: LateNight at 6:17 am (utc) on Nov. 28, 2003]

rfgdxm1

5:36 am on Nov 28, 2003 (gmt 0)

WebmasterWorld Senior Member 10+ Year Member



LateNight, my analysis shows this is being done with just a very select set of keywords. For example:

[google.com...]

[google.com...]

No evidence of stemming there. The way to test if a keyword is affected is to search it, and see if their is stemming evident in the SERPs. If there is no evidence of stemming in the SERPs for your keywords, then likely this couldn't explain a drop.

Krapulator

5:44 am on Nov 28, 2003 (gmt 0)

WebmasterWorld Senior Member 10+ Year Member



The stemming is very clever for place names - Australia also searches australian etc...a step forward in my opinion!

GoogleGuy

6:37 am on Nov 28, 2003 (gmt 0)

WebmasterWorld Senior Member 10+ Year Member



"The stemming is very clever for place names - Australia also searches australian etc...a step forward in my opinion!"

It's not just your standard stemming. :) There's some elegant stuff in there, and it will only get better over time. I'm excited about some of the newer tools we have to leverage linguistics and semantics to improve quality..

steveb

6:59 am on Nov 28, 2003 (gmt 0)

WebmasterWorld Senior Member 10+ Year Member



Suppose I want to search for singular but Google stems singular and plural together. What if I search for: singular widget -plural

Will singular be excluded too, leaving me with only "widget" results?

This seems like a bad thing to me. If I want to search for Australian I don't want Australia results cluttering everything up.

Pierre2003

10:33 am on Nov 28, 2003 (gmt 0)

10+ Year Member



This is a very nice thread. Let's see if I can 'summarize' the latest findings:

1. Florida fXXXed a lot of commercial sites.
2. Keyword stemming is found, but for only some keywords, especially commercial natured.
3. Link pop' and anchor text is still king, overulling titles & body.
4. The new filter, based on commercial terms/stemming or whatever makes it a much harder job to SEO as the margins from good SEO'ing to spamming are very close...
5. Based on the stemming findings - what are SEO's left to do in terms of keyword niches? I.E if In the past, we could find a niche keyword no one is trying to promote and get good positiones there quite easily - now, the stemming DB will rank other sites high for this keyword...
6. Does anyone have any advice on any pro-active 'what to do next'? (It seems everyone here is mixed up, confused and finding *amazing* findingd - but with no actual solutions...

sandalwood

6:38 pm on Nov 28, 2003 (gmt 0)

10+ Year Member



If anchor text and link are still kings, then we have to power down all our sites:)
Are you gonna have more anchor text and higher link pop than Y@hoo or Am@zon?

GoogleGuy

6:43 pm on Nov 28, 2003 (gmt 0)

WebmasterWorld Senior Member 10+ Year Member



steveb, I think you can use one of +word or "word" to get that exact word. I'll check to see if it's one or both; I think the plus sign definitely works though.

Robert Charlton

5:50 am on Nov 29, 2003 (gmt 0)

WebmasterWorld Administrator 10+ Year Member Top Contributors Of The Month



I'm excited about some of the newer tools we have to leverage linguistics and semantics to improve quality..

I hope this isn't off topic, but it seems like the right spot to mention it. I was playing around with the tilde operator... which previously produced results containing synonyms... and it now seems to be giving results that are much broader than they were before Florida.

Searching ~mattresses, for example, I see that the #4 result is for "Lava Beds National Monument." When I search ~kitchen, I see words like "food" highlighted. I can't compare these with prior results, but is the tilde operator in conjunction with Florida now doing something different from what it did before?

Just Guessing

2:07 pm on Nov 29, 2003 (gmt 0)

10+ Year Member



> When I search ~kitchen, I see words like "food" highlighted.

Interestingly the tilde synonyms for ~kitchen are different to the Adwords broad matching for kitchen. Adwords matches:

    kitchens
    appliances
    cabinet
    stainless
    cabinets
    ovens
    sinks
    sink
    granite
    butcher
    farmhouse
    stove
    franke
    grohe

and also suggests as a huge list of additional keywords, but it does not include food as a broad match or additional keyword, whereas the ~kitchen search does match on food.

GoogleGuy

8:20 pm on Nov 29, 2003 (gmt 0)

WebmasterWorld Senior Member 10+ Year Member



Hmm. I don't think the tilde results have changed recently. Robert Charlton, you raise a good point in that the tilde operator gave an early glimpse that Google was starting to learn more about word meanings and semantics.

Toasted

9:58 pm on Nov 29, 2003 (gmt 0)

10+ Year Member



Don't know about you guys, but I'm seeing some weird results with the new stemming on Google... not sure whether it's just very crude, or whether I'm getting some weird caching or data center jumping happening.

Anyway, does anyone see anything different to this:

1. Search for 'blue widget' - also highligts 'widgets', 'widgetgizmo' etc.
2. Search for 'widgetgizmo' - no stemming at all.
3. Search for 'blue widgets' - no stemming (i.e. doesn't highlight 'widget').

If this is indicative of how it works, then it is reasonable to assume that Google has a static list of stemmed words - this list is far from complete and all stemming is one way only (i.e. widget = widgetgizmo, but widgetgizmo!= widget... unless that is also added as a separate entry).

Can anyone see anything different? A bit odd don't you think?

Further more - can anyone find an example of stemming that doesn't also correspond with one of the major changes in serps?

flicker

12:42 am on Nov 30, 2003 (gmt 0)

10+ Year Member



I mentioned this way the heck back in one of those locked update threads, but I've been seeing signs that Google has correctly realized that a common pair of two-word phrases are synonyms. This has really strengthened some of the search results I've seen. I hope this trend continues... though an exact match is still the best result, I'm so much happier seeing a few good "used couches" results returned in my "used sofas" search rather than extraneous porn spam with "sofa" anchor text, or whatever. (-:

I could just be imagining it, of course, but I hope I'm not. (-:

Just Guessing

10:57 am on Nov 30, 2003 (gmt 0)

10+ Year Member



At the moment I am not seeing word stemming for single keyword searches, not even plurals.

But I am seeing word stemming for the same keyword in some multiple keyword searches but not others including the same keyword.

Specifically

Widget does not match Widgets
Blue Widget matches Blue Widgets
Red Widget does not match Red Widgets

All very, very strange!

I don't see synonyms unless I use the Tilde ~keyword search.

Pierre2003

11:56 am on Nov 30, 2003 (gmt 0)

10+ Year Member



try searching your phrases then add -baysianspamfilter
AmaZING! it works for onbly some keywords!

Just Guessing

6:00 pm on Nov 30, 2003 (gmt 0)

10+ Year Member



Yes both Word Stemming and the "Commercial" Filter go away with -sdhfk, but I am sure the filter is definitely not just different results from Word Stemming.

egomaniac

1:44 am on Dec 1, 2003 (gmt 0)

10+ Year Member



Wow those negative searches are pretty trippy.

Just_Guessing, what does "-sdhfk" mean? (I get the part about the minus sign excludes it from the search)

Pierre2003, what does "-baysianspamfilter" mean? I've never heard of this.

When doing both of these searches, the results were nearly identical to pre-Florida!

Pierre2003

10:00 am on Dec 1, 2003 (gmt 0)

10+ Year Member



Exactly egomaniac!
This is a trick to see the pre-florida results.
This 52 message thread spans 2 pages: 52