Chiedere al guru seo

First you need to clarify.. I didn't say LSI was behind it exactly. I have never been one to claim Google implements LSI...but rather Google has implemented an LSI-like approach, which may sound like the same thing but it's not.

I initially called what I saw in '03 "latent semantic <em>imaging</em>" or LSI before I knew LSI was being used in the SEO community. A latent image is a captured image which is not evident, except through investigation. An exposed photographic image is a <em>latent image</em> until the film was developed. In SEO, if my page talked about X and Y but not Z, yet Google ranked it for Z even though Z is not a synonym for X or Y, and the anchor text did not include X nor Y, then Google was holding a <em>latent (semantic) image</em> of my page, considering it to be about Z. The cause is Google's implementation of a semantic relevancy weighting system which I believe is based on practical experience researchng latent semantic imaging on the Google index. This was tested in '03-04 and abused heavily by some SEO types. I believe it is now back, well-refined.

If you can make this a debate of semantics vs. links, I'm game. But if you state is as LSI vs links, I plead no contest and respectfully decline. Oh, and to make it even more confusing, there is a whole web community talking about the semantic web now, which is also quite different (although, sadly, related).

>I have never been one to claim Google implements LSI<

fair enough. It is easy for me to get confused when it comes to the acronym LSI used in an SEO context. I first
became aware of the term from Edel Garcia http://www.miislita.com/. He and I became friends during the
SearchKing lawsuit and he was talking it up back then. But he was talking about Latent Semantic INDEXING. This
is the first time I have heard the term LS IMAGING and now I have to get my head wrapped around THAT :(

To be honest I never had much affinity with the whole LSI thing mostly because I couldn't see the direct results
like I could with links. So, taking the path of least resistance, my focus was always much more targeted to the
right links from the right site saying the right thing. I dabbled with the text saying x and y and placing for
z but while I did see some positive indications, I wasn't all that impressed with coming up #7 for fruit when
the page talked about bananas and apples. I was never able to feel like I was really in control of the results
like I was when dealing with links.


>but rather Google has implemented an LSI-like approach<
are you referring to an LSI like approach as imaging or indexing? Not that I even REALLY understand the
difference but it does seem to be a finer point.

>This was tested in '03-04 and abused heavily by some SEO types.<
I still don't see how you can abuse LSI beyond a gut feeling without having some kind of list showing the
semantic relationship between all terms. Without that it still seems to be a shot in the dark to target fruit
but write about bananas and apples. Why wouldn't you come up for trees or peelings?

The reason I've all of a sudden developed enough interest in the topic to be willing to debate someone of your
caliber of expertise is that I AM starting to see some type of latent semantics in anchor text, (I think), where
the anchor text of 10 strong links would say bananas and another 10 say apples yet the target site starts moving
up for fruit. ????:0 ???

I'd like to understand it better but I still it is the links that move a site for target terms unless we're
talking about just the long tail stuff.

Good discussion guys.

There are potentially so many different things going on with Google right now, that trying to pinpoint any one thing might confuse matters.

1. Better indexing by use a compression/decompression system that allows Google's index to hold more information within less space, and simultaneously allows better phrase indexing, the inclusion of stop words within results, and possibly the use of a multi-staged query system that allows for stemming of query terms as well as query expansion.

2. A phrase based indexing system that uses co-occurence of phrases to maybe achive some results like what John talks about with what he is calling latent sematic imaging, to rerank results (If phrases x and y often appear with phrase z, then a document that contains phrases x and y might have something to do with phrase z).

3. The use of a very large body of information contained within the queries, searching behavior, and browsing behavior of people on the web could be useful in understanding what people find interesting, or relevant, or useful.

I'm sorry but I honestly do not see any value in going further in a public forum like this.

Bill I love your editorials on the patent activity - it adds great value and it takes a certain
mind to pull it off. But practically, if I start to say "check out how this analysis tool
reveals this about Google's process, so when you do this or that you see this impact on your ranking"
I'm just pointing a community of very ambitious determined and bright entrepreneurs at a target.

As much as I'd love to discuss this with you guys (because I know I would learn as well as enjoy the
back and forth) it won't bring any joy to the business end that funds my endeavors. Things move too
fast for the academic approach when you've got limited resources to throw into R&D.

As much as I hate to admit it, you have a valid point john. Like you said I would LOVE to discuss this publicly
in the hopes of resolving even one more goofy debate about links vs content but I doubt we would do much to
change the minds of those who don't know and alter the fortunes of the ones that do at our own expense.