Florida #2 ?

[I, Brian]

Ah, sorry - I misread your previous post.

What I meant is that for "Florida" results my personal interpretation is that an expert system - such as Hilltop, or TSPR, or something else - makes the initial selection of documents.

An expert system, so far as I can understand it, can only work on a limited set of search terms. The update over the past weekend could have been a simple widening of this process.

However, even for terms outside of the Florida affect, I'm seeing a general devaluing of internal links - especially forum links. The simplest culprit for this would be something like LocalRank.

So I see an expert system - such as Hilltop - as being primarily involved within Florida effected results - - - with LocalRank being tightened on returns outside of the Hilltop set of terms.

So I see an argument for both being applied in different ways - which is how I actually read your post above.

[glengara]

As both mention " non-affiliated "pages, the future for internal links and link text is bleak all right.

[compar]

Let's get some history here. Hilltop is a research paper published by:

Krishna Bharat
Compaq, Systems Research Center, Palo Alto, CA 94301
(Current Address: Google Inc., 2400 Bayshore Parkway,
Mountain View, CA 94043)
krishna@google.com
George A. Mihaila
Department of Computer Science
University of Toronto.
georgem@cs.toronto.edu

At some point in time after the release of the Hilltop paper Google hired Krishna Bharat. And subsequent to that they filed the Patent for Local Rank. Here is what the Patent says:

Inventors: Bharat; Krishna (Santa Clara, CA)
Assignee: Google, Inc. (Mountain View, CA)
Appl. No.: 771677
Filed: January 30, 2001

Now why would Google bother hiring the person who did the Hilltop research and then file a patent of a similar methodology if they intended to use Hilltop? The fact is that they are not using Hilltop. They don't have a patented on it. If they are using anything, they are using Local Rank which is a closely parallel technology for determining page relevance based on the linking schema of relevant pages.

Hilltop was an attractive idea. Google likes to patent technologies for all the obvious reasons. So they hired one of the authors of the Hilltop paper and modified the approach slightly and patented it.

If they were using both Hilltop and Local Rank it would require two additional rankings of the results rather than one. That just doesn't make any sense. It also doesn't make any sense that they would patent a technology for doing essentially the same thing and then park the patent and use an unpatented procedure.

[I, Brian]

I am quite aware of the issue of Krishna Barat's relationship to Google, and also the fact that Google applied for the LocalRank patent. You don't need to state the obvious.

The point about this discussion, really, is whether LocalRank can effectively define an "expert system". I still have difficulty in sharing your own interpretation. You could be completely right, or completely wrong. But, to be honest, there are still a lot of uncertainties, and I'm trying to expect the unexpected.

Whether Hilltop itself is being used is certainly a contentious issue (it is an old paper, and would almost certainly have to be quite revised) - and there's some steam behind the theory that it is Topic Sensitive PageRank being applied.

Either way, there seems to be some form of "expert" system in place. What exact form that criteria relies upon is up for debate, but there are no "complete proofs" to back any position as yet. That is quite possibly the best common ground in SEO at the moment - though the details are bound to be discussed and debated.

[glengara]

*You seem to think that there is something magic about "expert". There isn't. A page just gets to be expert by virtue of its content and the people who are linking to it. *

I think you may have read that wrong, Compar, here's what they say:

*We define an expert page as a page that is about a certain topic and has links to many non-affiliated pages on that topic.*

[glengara]

<added> As to the two being identical, AFAIK you can't get a patent for something that has already be published, so either the patent was applied for at the same time as publishing Hilltop, or they are indeed different.

[compar]

and there's some steam behind the theory that it is Topic Sensitive PageRank being applied.

And I don't believe it is TSPR either. All of these tolls try to do similar things including applied semantics. They all try to improve the relevance of the SERP by examining the linking structure in a query sensitive way.

Previously anchor text was probably the only measure of relevance.

PageRank of course has never had anything to do with relevance directly. It is a sharing of numeric rank and is independent of the query or search term used or the content of the page from which the inbound link comes.

But back to TSPR. I doubt that Google is using anything on which the do not hold the patent. Therefore it is my speculation that new relevance measure is being done by Local Rank.

[juliany]

IMO, an experte site is one with a perfectly defined theme. Please discuss this in my post "themes" in this Google Forum.
I think it's the answer to this mayhem.

Julian Yanover.

[jonsimons]

this 'update' happend almost a week ago now.. does this mean that its more than likely a permanent change, and so we should start thinking about changing our sites?

[I, Brian]

Originally Posted by compar

Previously anchor text was probably the only measure of relevance.

PageRank of course has never had anything to do with relevance directly. It is a sharing of numeric rank and is independent of the query or search term used

compar, I would be interested to read your justification of that position.

After all, anchor text is a powerful tool, but all sorts of factors have always been involved in the ranking of pages on Google.

And PR most certainly has been directly used for the ranking of pages - it's basic SEO and SE theory:

Anatomy of a Search Engine

These maps allow rapid calculation of a web page's "PageRank", an objective measure of its citation importance that corresponds well with people's subjective idea of importance. Because of this correspondence, PageRank is an excellent way to prioritize the results of web keyword searches.

[I, Brian]

Originally Posted by jonsimons

this 'update' happend almost a week ago now.. does this mean that its more than likely a permanent change, and so we should start thinking about changing our sites?

Probably a permanent change in terms of a different form of algo being used - but the current SERPs should be considered subject to refinement - initial Florida results changed quite dramatically over the weeks that followed.

[Emancipator]

Originally Posted by compar

Previously anchor text was probably the only measure of relevance.
PageRank of course has never had anything to do with relevance directly. It is a sharing of numeric rank and is independent of the query or search term used

Originally Posted by I, Brian

And PR most certainly has been directly used for the ranking of pages - it's basic SEO and SE theory:

Brian I dont think you are seeing what Compar is saying. Page Rank does NOT determine your relevance. Page rank is a popularity contest. Relevance is not the same thing. How can a site with no relevancy but high page rank show for any term? You will need relevancy even if you have page rank.

Topical Page Rank 'IS' a totally different conversation but until somebody can show me proof of use and calculating it, I leave it like the "god" factor. Nobody sees it, but we know its there, so lets leave that one be.

Brian i like that for every statement you make you reference a source, good show.

[glengara]

After some re-reading, and if I understood it correctly, the main difference between Hilltop and The Patent is:

Patent
Initial ranking based on on-page factors, followed by a ranking based on off-page factors from within the initial results.
The combination of the two provides the new TSPR.

Hilltop

Initial search for topical "experts", followed by calculating the number and relevance of "expert" links pointing to a particular page.
Also provides a TSPR.

So while the object of the two is similar the "ways and means" are not, and, IMO, they are complimentary and would work best run in parallel.

[I, Brian]

Originally Posted by Emancipator

Brian I dont think you are seeing what Compar is saying. Page Rank does NOT determine your relevance. Page rank is a popularity contest. Relevance is not the same thing. How can a site with no relevancy but high page rank show for any term? You will need relevancy even if you have page rank.

I think myself and compar may be using very different applications of the word "relevant".

Certainly I agree with bob that PR is not related to "theme" or "topic". I think that compar is trying to make that particular distinction.

For myself, I've been meaning "relevance" within the set of returned documents for a search term.

So PR by itself will not help a page on "widgets" rank highly for a search on "fruit" (ie, PR is not relevant) - but within the set of documents returned for "fruit", PR is involved in ranking the returns (ie, PR is relevant).

I think our entire discussion is based on a misunderstanding of this differing personal application of the term, which means we each perceive strange errors in the rationale of the other.

Semantics again, I guess.

[Emancipator]

I think myself and compar may be using very different applications of the word "relevant". ertainly I agree with bob that PR is not related to "theme" or "topic". I think that compar is trying to make that particular distinction.

For myself, I've been meaning "relevance" within the set of returned documents for a search term.

So PR by itself will not help a page on "widgets" rank highly for a search on "fruit" (ie, PR is not relevant) - but within the set of documents returned for "fruit", PR is involved in ranking the returns (ie, PR is relevant).

Page rank is simply popularity of a site. I think the problem is compar is talking about relevance of page rank and you are talking about some other form of relevance which I am not sure what it has to do with the topic. Page Rank in itself does nothing to improve search relevance which is what we are talking about here, starting with compars post. I agree with compar 100%. Although page rank does affect your ranking it does not play into the scale of relevance.

Small clarification of my mention of "page rank". I mean page rank that everyone stares at on the Google Toolbar, not actual Topical Page Rank that I mentioned earlier, which does play into relevance.

[I, Brian]

"Although page rank does affect your ranking it does not play into the scale of relevance."

I do quite agree with bob about the issue of PR not being a selection criteria - when I used the term "relevant" I meant simply that PR is normally seen as being used (to some extent) in ranking the returned set of documents related to that search term. Perhaps within all this mis-understanding is my being sloppy with definitions of the English language?

[Emancipator]

lol.... At least you took the time to clear it up. I was staying out of the conversation until it honestly came across as you saying page rank was relevance, which I was PRETTY darn sure was not what you meant.

You are using relevance as a factor scale, not as a search relevance scale. I gotcha.. makes sense, hopefully makes sense to everyone else reading this wondering wth you and I are yammering about

Always like the fact that you come back to clarify your points without biting heads off I,Brian.

[dpam]

Originally Posted by JohnScott

Go do a search for "cheap ******" and check out all the factors. You'll see the right way to do SEO.

John could you summarize what it is you see. A brief look shows me (for the current #1 anyway) marginal PageRank and lots of garden-variety link exchanges. What is the trick sending that guy to the top?

[Rob]

Originally Posted by dpam

John could you summarize what it is you see. A brief look shows me (for the current #1 anyway) marginal PageRank and lots of garden-variety link exchanges. What is the trick sending that guy to the top?

Inbound links.

[compar]

compar, I would be interested to read your justification of that position.

The first thing we obviously need is to agree on the use of the word relevant or relevancy. Google say that what they are always trying to do is deliver the pages that they think are most relevant to the search term. I take it from that they mean the pages that they think will best provide the information the searcher is looking for or interested in.

So when I use relevance in any discussion about Google, or search engines in general, I confine the word to this limited definition.

Now I have searched and cannot find it, but I saw someone explain the other day -- and I thought it was in the Topic Sensitive PageRank paper -- that the original concept of PageRank was that links would normally be used between sites of similar content or themes, and while PR was simply a mathematical calculation if the assumption about the nature of links had held true it would have also been an indirect measure of relevance.

However we know that as soon as webmasters figured out the importance of links, this assumption went out the window. So today PR is a sharing of the numeric value of link pages and Google has been force to look for something else to denote topical or query related relvance of a link

For a long time to the best of anyone's knowledge this was the anchor text. It was the anchor text that suggested what the target page was about. If someone was searching for "cheap ******", and Google saw a whole bunch of links to my page with "cheap ******" in the anchor text their assumption was that my site must contain information about or be relevant to the search term "cheap ******".

They can't do this any more from the PR number because their original assumption about all links being from like relevant pages is not valid.

So PageRank has nothing to do with relevancy. And it is my suspicion that it is becoming less and less a factor in the decision about which pages to put in the SERPs.

I think Google is trying harder and harder to measure the topic or theme relevancy of links. Whether this is being done with Applied Semantics, TSPR, Hilltop, Local Rank, Stemming, or anchor text, or some combination of all these things is the unknown. But Google is looking for relevancy and PageRank is not a measure of this.