| 3:35 pm on Oct 3, 2002 (gmt 0)|
Not that I can see. As I've mentioned, I would be careful building too many small sites that you try to leverage the linking for a main site. I'd much rather see one larger main site that stands on it's own.
Insite linking (links within the same site) across your relevant content has never been more important. Share that pr and content relationships between your pages:
Fruit links to Oranges that links to Juice that links to Apples that links to Pie that links to Cake that links to Icing that links to Sugars that links to Sweetners...
| 3:40 pm on Oct 3, 2002 (gmt 0)|
>Fruit links to Oranges that links to Juice that links to Apples that links to Pie that links to Cake that links to Icing that links to Sugars that links to Sweetners...
So it would be more beneficial to have 2 or 3 (or even 1?) VERY relevant links on the page than to have say a template with your 8 main catagories:
Fruits ¦ Oranges ¦ Juices ¦ Apples ¦
Pies ¦ Cakes ¦ Icing ¦ Sugar ¦Sweeteners
at the bottom of EVERY page?
I am still somewhat baffled on Google's use of 'themes' I guess.
| 4:20 pm on Oct 3, 2002 (gmt 0)|
You've probably read some of the recent discussions on the possible use of "context". I say possible because no one has yet quite figured it out. The working theory is that the page title of the linking page helps define the destination page context.
The easiest way to understand how context can work is by going to the Google labs utilities and trying a few examples in the Google Sets. If you put in Apples, Oranges, and Pears, it prints out a list of fruit. Given that, we know that Google knows completely how to form relationships (umm, some times called Term Vectors) between individual words.
Extrapolating that to "context", we could look at linking text, page text, and title text of any linking page. If "oranges" links to "apples" with the link text "fruit" and the title text "oranges", then the "sets" detector can see there is an appropriate relationship between the two pages.
That helps Google to know that it is a quality link between two pages. It knows that the oranges page linking to the apples pages is a good relationship and the pr should be passed onward. eg: its a theme.
In regard to menu items. Google appears to be excellent at detecting like templates. The most obvious way to see that is via Google News. That is all automated. It is detecting templates, removing the template text and left with the results content of a page. From there, they can determine if two stories are the same story and instead of listing it twice, they list it as a sublink (all the ap and reuters stories are id's that way). That is the same way a dupe detector works.
Template or menu links in your site are not a problem for Google to see and understand. If they can see the content on a page and strip away the template, they can spot dupe content between sites using vastly different menu systems. Only Google knows how they are treated of course, but most of us are working under the theory that they can't hurt you and little of that pr from menu links is probably passed between pages. Much more important are the links buried within content. I think this was one of the major changes here in 2002 and somewhat this current update.
|brotherhood of LAN|
| 4:34 pm on Oct 3, 2002 (gmt 0)|
|Much more important are the links buried within content |
Makes good sense to me.
Goes in good stead with what Brett wrote about theme pyramids [searchengineworld.com] and what ciml wrote on theme pyramids and google [webmasterworld.com] alongside the "in 12 months alone" text. Links that are "in" the content are most likely to be the most on topic so more emphasis could *perhaps* be placed on them.
All the theories that are flying about in regards to the update hint that PR for sale and google bombing could be cured by what Brett writes above. When searching for "microsoft" there would be no real 'meaning' of "go to hell" when searching for microsoft, because upon google visiting their site - they would see that all things microsoft have nothing to do with hell :)
I just wonder how google goes about making these "themes". Would they define a theme or use some sort of AI or clever math to assume a "theme" from the webpages it sees. i.e. if enough pages say "microsoft" and "hell" together then their is a relation for those queries.
I can see why they call them theme pyramids, the problem is from a webmasters point of view guessing google ...its more like a stack of cards from here :)
| 5:15 pm on Oct 3, 2002 (gmt 0)|
|they would see that all things microsoft have nothing to do with hell |
don' want to go off topic but i would say "they would see that NOT all things microsoft have SOMETHING to do with hell" ... ;)
However, i'm a big fan of the 12-Month and the pyramide theory - it works in real life!
|if enough pages say "microsoft" and "hell" together then their is a relation for those queries |
last off topic: probably ... ;)
| 5:20 pm on Oct 3, 2002 (gmt 0)|
That microsoft search results only works because it is a phrase. When you search for a phrase, google has to match the entire phrase. you aren't searching for "hell", you are searching for "go to hell" and it is treated almost as if it were just one word. Context would be difficult to inject into that type of search.
| 5:29 pm on Oct 3, 2002 (gmt 0)|
Wow...I didn't know about google labs....what a find....you can bet I'll be playing with sets for a while.....
In regards to templates...
>Template or menu links in your site are not a problem for Google to see and understand. If they can see the content on a page and strip away the template, they can spot dupe content between sites using vastly different menu systems.
Am I correct in assuming then, that if I randomize my link text at the bottom of pages, it may have more weight?
From the looks of this statement, it assumes that google may lower the weight of link text throughout a site if it is deemed to be 'template based' (repeating in the same form on multiple pages).....my pages are nearly all template based, but I'm wondering now if I may have better luck from creating LOTS of different templates to retain some ease for updating, but at the same time boosting the weight of my link text a bit.
| 5:41 pm on Oct 3, 2002 (gmt 0)|
I'm finding results for 3/4 word searches truly awful.
If this continues, would it not greatly impact on the effectiveness of the themed pyramid structure?
| 7:12 pm on Oct 3, 2002 (gmt 0)|
It's still too early to be confident of what the recent changes are precisely, but I'd like to mention that certain aspects of the 'Google Pyramid' post may be considerably less relevant now. Other aspects, I still expect to be relevant some time in the future.
> ...The working theory is that the page title of the linking page helps define the destination page context.
I expect that to become important, but I can't see it yet. Google may even skip to 'titles of the pages that link to the pages that link to you' if they can do that efficiently.
> I just wonder how google goes about making these "themes".
Then you're ahead of me. I wonder how Google will go about making themes. Sorry if that's what you mean!
| 7:24 pm on Oct 3, 2002 (gmt 0)|
Don't forget Google wish list ... er ... suggestions.
*Detecting common templates in pages, and separating out the common structure from the individual content. (as Brett hinted at with the News section)
*Classifying links on a page.
*Detecting pages that are near-duplicates of one another.
*Clustering pages by topic or type.
I think I read a paper by one of the Google engineers (formerly at AltaVista?) about using ODP data to judge what category a page is in.
|brotherhood of LAN|
| 7:25 pm on Oct 3, 2002 (gmt 0)|
>what I mean
So you think there is no element of "theme" already in the algorithm?
Be it title of a page with links, a DMOZ category, DMOZ category of link page.....I would have thought that Google would apply something that de-emphasizes PR and emphasizes "themes" more?
I don't have an idea of what scales Google works on but I'd like to think they've got a few ideas in the hatch to enforce on-topic linkage - by one way or another recognising the subject area of a page/site.
| 2:21 am on Oct 4, 2002 (gmt 0)|
About Google labs....
Try the word astrology and see how many irrelevant matchs it offers.....Sheesh, if I used links to sites like that I would get penalized for irrelevant linking,,,,,CRAP! thought I found some good help.
| 6:06 am on Oct 4, 2002 (gmt 0)|
try more than one word: for example:
and Google sets results become far more relevant.
So when Google is searching for something [webmasterworld.com] to make context, Google could try to match anchortext (astrology) with e.g. title/description etc (zodiac) of linked to page. (just one example).
I would say all this matching must be expensive computational wise, and much more difficult for foreign languages. (e.g. Google sets does not work for languages other than English).
On the other hand, most over-optimisation will probably occur in English. ;)
I would agree with slud that Google probably looks at DMOZ first for looking for "theme related phrases/words". At the moment they might even only use synonyms, dictionary description words plus exact word matches (single/plural), as integrating real theming related words such as e.g. Google sets or signatures is a bit to advanced for Google yet. [webmasterworld.com]
(the only way my father could push me for extra work, was saying something was way too difficult for me :))
| 6:28 am on Oct 4, 2002 (gmt 0)|
I feel like I've been spitting in the wind all this time.
Google is definitely putting more emphasis on inbound links, title tags, etc. for defining page relevance to a particular search term...
and less emphasis on what YOU structured/optimized the page for.
It makes sense. I'm sorry for being so harsh on you Google!
Google... Please don't penalize me for my freedom of speech! Look into your heart!
| 6:34 am on Oct 4, 2002 (gmt 0)|
I agree with your advice to ann, vitaplease, with regard to context searches and google's own way of establishing a contextual link between search terms. I believe you will find that the directory element of google is particularly influencial when it comes to which sites are selected for appearance against context.
Our own site is a typical example. Originally we were very much into 2nd hand equipment so we went into the "used" section of the directory. (We've never been able to get out of it!) and now we still appear high in searches where for example the terms are "pc" and "upgrade" then google automatically sees used computers (the directory we're listed in) and throws us into the results.
The directory listing is definitely a big part of google's context algorythm.
Hope this helps,
| 7:24 am on Oct 4, 2002 (gmt 0)|
|I'd much rather see one larger main site that stands on it's own - Brett_Tabke |
First on this I so agree Brett. And I bet there would be a lot less frazzled nerves each Google update, if folks would spend more time building on the strategy you offered us with this. Combine that with a good theme development for a winning hit, provided the content is there and fresh. I think every industry, market, even down to specific keywords will all require a tweak here and there but that’s the little bit of magic we all bring to the mix.
Now that’s not to say that you can’t be incredibly successful multiplying this plan for all the various ideas and projects you may be interested in and create an empire of themed sites. If there happens to be legitimate reasons to link one site to another then that shouldn’t be a problem either. A really large and appealing site can become a hub if necessary or desired.
I guess my opinion is this is a great template to work from. Then, I think it comes down to content. Keeping it fresh, informative and attractive to both search and visitor.
I spend hours in research each day, studying different industries. I guess what I was trying to express on the first pass through in the Google update thread is much of what Brett is explaining here. I have noticed trends developing over time. It’s as if a change is made and with each update this original change builds. It takes out a few sites with each update and then hits a critical mass and the changes are more obvious. People think they’ve happened because of a sudden change when in reality the change occurs over months.
|The working theory is that the page title of the linking page helps define the destination page context. - Brett_Tabke |
|I expect that to become important, but I can't see it yet. Google may even skip to 'titles of the pages that link to the pages that link to you' if they can do that efficiently. - ciml |
I have seen this pattern, at least Brett’s suggestion here to a theory. I’m going to go back and apply your suggesting, ciml – to my research because I wasn’t considering that.
The only part of the “'Google Pyramid' post” that I would expand on is the internal linking. I do more cross linking with my themes then vertical linking and my themed sites are more like a planetary system then a pyramid, which is due to my use of canonicals.
|Much more important are the links buried within content |
…ah, Brett giving the bank away…
| 1:24 pm on Oct 4, 2002 (gmt 0)|
I don't buy into the themeing thing. I qualitatively I don't think you'd end up with better search results for the user by employing such a technique.
This update smells more of knob turning than re-engineering.
| 2:07 pm on Oct 4, 2002 (gmt 0)|
|I would say all this matching must be .. much more difficult for foreign languages. (e.g. Google sets does not work for languages other than English). - vitaplease |
why would that be? for google a word is just a mix of characters, regardless of language. apparently google just needs a very large sample to create those sets as they work in big languages only (i tried german, spanish, french and italian) but not so in smaller ones (i tried dutch and it didn't even work on the main sets such as weekdays and numbers).
| 7:45 am on Oct 7, 2002 (gmt 0)|
|apparently google just needs a very large sample to create those sets as they work in big languages only (i tried german, spanish, french.. |
I agree, but even for French and German the sample does not seem to be large enough.
If I try in German: "Anorexie" and "hungrig" I get no results
If I try in French: "odeur" and "chloroforme" I get no results
the same queries in English give satisfactory results.
Also if Google would do some matching with DMOZ categories, they would find the non English languages are not represented well enough for good comparison.
If Google uses their own Glossary as an input/check, then again alas, they also only use (show) English on-line glossaries. Most probably again the online amount of foreign language glossaries are too limited.
I am sure Google has not committed themselves to other languages yet (the least they could do is import a dictionary/glossary/synonyms lists etc. - but I guess that would mean copyright issues as well as soon as quality dictionaries are used).
for those interested in some speculations on how Google sets works:
some thoughts [groups.google.com].
Same goes for Altavista Prisma by the way, as far as I can see relational terms only show up for English queries.