Sometimes the structure created by trying to “silo” your site can be beneficial, and other times it might be improved if you remove your silo blinders and build the site structure with internal linking and PageRank distribution in mind.

Some points we already agreed on, others I feel you might have leaped to conclusions about what we meant.

a) You’re correct – LSI is possible, just not feasible under its original technical implementation. From a search engine’s perspective, each page added would not only expand the complexity of the index/analysis – but also require the data to be re-analyzed. And this would require an exponentially growing number of computing resources.

This is the first and probably most obvious flaw in LSI for search engines.

As I mentioned in the post, it’s very likely that Google has developed something similar – but then we’re not technically talking about LSI anymore.

b) You’re correct, 6 years ago Google bought out a company who had done a lot of research around LSI – but if you watch their Tech Talk videos since, they appear to have found a similar, faster(?) technique (as I mentioned).

As you’d know already, Google has been particularly silent on LSI in recent times – but they’ve released plenty of information on what they refer to as “The Semantic Web”. (I think I mentioned this in the post too?)

c) I couldn’t find where I referred to LSI as new. Let me know if I’ve missed it though. You’re right – the concept is over 20 years old.

d) You’re technically correct that people don’t “do” LSI – however, I think you might have missed my paragraph on this.

Because of a lot of SEO mis-information LSI has become a defacto name for particular types of keyword relationships (see my paragraph before the title “Nuts and Bolts”).

e) Re: LSI is not about structuring – refer to above.

We’ve since begun referring to it as “semantic site structuring” as we feel this is a more accurate description than “LSI”.

(We want to stop the bastardization of the term too. And if we use a more accurate, descriptive term then perhaps people will stop wrongly referring to it as LSI.)

Our own research, and others too, suggests that with the right site structure a site can achieve rankings faster, with less effort, and fewer links. (I believe I saw some others posting comments here who had independently come up with the same conclusions.)

However, the reason why I’m unsure about. My best theory is that it’s to do with internal site linking from relevant pages.

I knew I’d find LSI controversy from this post

Thanks for keeping us on our toes

We do read the research papers, we have done our own research into LSI, we even implemented our own version of it as part of a research project. (Fascinating, by the way. I highly recommend trying it out for yourself. I think you’d get a real kick out of it!)

If we’ve drawn the wrong conclusions from the data though, I want to know. (My ego can take it. If I’m proven wrong, I’ll be richer for it in the long term )

Brent

Brent, as Arik (and others?) has quite rightly said already, LSI is NEITHER a Myth NOR a lie. It is patently ludicrous for ANYONE to try and state or imply otherwise.

The drift of many of the comments herein appear to indicate a general belief in one or more of the following:

a) LSI is not possible (CR*P!) – Yeah! You’re a believer in the “Men never landed on the moon” conspiracy theory too I bet.

b) Google is NOT using LSI (more CR*P!) – Err! Ever heard of Adsense? Of course they’re using it. See: http://bit.ly/CKz3f for Google press release – 6 YEARS OLD!! Read CAREFULLY the third paragraph. Pay special attention to that word (a Trade Name) in the middle of the second line. Recognise it? Now you know where it came from – NOT Google.

c) LSI is a “new”, untested, early, Alpha or Beta-level technology (utterly farcical, even more unbelievable CR*P!) – Good God people! LSI as both concept and deliverable technology is WELL over 20 years old for crying out loud! 20 years(!!) in an industry (IT/Data Mining) where 1 year is equivalent to 5 or 10 in many other industries! People were doing LSI for real at least 10 to 12 years before Google started using it in anger.

I remember discussing LSI at some length with a senior AltaVista VP at DEC (DEC = Digital Equipment Corp. inventor, developer & original owner of AltaVista) over 17 years ago, about two years before I left DEC, having spent almost 16 years there, from senior engineer to UK & European Sales & Marketing.

d) People are “doing LSI” on their websites. What? Are you NUTS? So much CR*P you could overflow four sewage farms! Let’s be CRYSTAL clear here – LSI is “done” BY Google (and other SE’s like Yahoo!) on YOUR site’s data & meta data NOT the other way around.

YOU do “SEO” (or you should) on your site, albeit now, hopefully, with LSI & its ramifications firmly in mind. NO mention of LSA though I see; THAT’s what we SHOULD be optimising for. After all, its the ANALYSIS of what’s been indexed (By Google et al.) that’s important NOT the indexing per se.

Ref the link below to see what LSA is about and how LSI relates directly to it. It’s the analysis that produces (for Google) the theme of your web site and YOUR job is to try your damndest to ensure that YOUR (intended & designed) theme MATCHES what Google (or any other SE) concludes from its LSA.

e) LSI is all about “structure” – HORSE MANURE!!! – NOTHING could be further from the truth. Go research REAL info on LSI (scientific & learned papers) as suggested below.

The bottom line(s) here is(are):

1. CONTENT (highly descriptive, detailed, linguistically-rich, inter-related phraseology content AND META DATA CONTENT) is KING!

2. The THEMATIC “appearance” of a site to Google’s eyes will ALWAYS be more important than the site’s structure. However, good structure (and a decently organised site-map) will always ease the job for both robots & human visitors and hence can only enhance the value (ranking?) Google puts on it (LSI or NO LSI).

3. The theme of your SITE overall is as important (potentially) as the theme of your individual PAGES. Maintaining consistent, albeit differing, theme-related phrases & context across pages & the site is key.

If a silo structure approach helps achieve this then fine. But it is the consistent and related theming that is most relevant NOT the structure per se. If this were NOT true then marketers who stress their lack of doing (classic) SEO but showing HIGH and MONOPOLISING positions in the SERPS simply would not achieve those positions.

Want to get the REAL skinny on LSI? Undertake a search for “Latent Semantic Indexing Scientific Papers” (not for LSI on its own as that term appears to have been hijacked by “Know Little, Guess More, Spout a Lot” SEO poseurs!! – Over 23 pages of SERPS for Latent Semantic Indexing and barely a single reference to a learned paper!?

Go do some basic, fundamental research (it’s called RE-search for a very sound and appropriate reason). You may be surprised by what you learn from the “bleeding edge” brigade of REAL researchers on LSI. Be sure to check out the dates of some of those papers too (early 90’s and earlier).

I highly recommend those who have further interest in understanding how LSI & LSA can be expected to operate, with regards to SEO at a simplistic level, go and visit this site:
http://bit.ly/BrvwQ – I DO take issue with their titles on slides 11 & 12 however – they SHOULD read “Implementing Theming On Your Website”.

BTW before someone slags me off for this rant(?), I have absolutely NO problem with empirically tested/proven SEO techniques – I’m a graduate engineer after all and unlike typical scientists (who don’t seem to believe ANYTHING is possible before they’ve developed and proven their theory for it) I’ve always had a philosophy of “If it works, use it”.

The techniques outlined above have been proven clearly to improve results, so use them. But PLEASE, PLEASE stop saying “I’m doing LSI” … GARBAGE! You SHOULD be doing SEO. Weren’t you always? But you’re definitely NOT “doing” LSI (well not unless you’re developing a rival to Google or Yahoo!)

Let’s stop making farcical leaps of logic and drawing patently false conclusions equivalent to: “Cox’s Orange Pippins are apples” ergo “Apples are Cox’s Orange Pippins”… NOT!

Best regards.