Search Engine Watch
SEO News

Go Back   Search Engine Watch Forums > General Search Issues > Search Technology & Relevancy
FAQ Members List Calendar Forum Search Today's Posts Mark Forums Read

Reply
 
Thread Tools
Old 08-25-2004   #1
rustybrick
 
rustybrick's Avatar
 
Join Date: Jun 2004
Location: New York, USA
Posts: 2,810
rustybrick has much to be proud ofrustybrick has much to be proud ofrustybrick has much to be proud ofrustybrick has much to be proud ofrustybrick has much to be proud ofrustybrick has much to be proud ofrustybrick has much to be proud ofrustybrick has much to be proud of
Themed Sites Level of Importance in Ranking Algorithm - The Theory

I have had the honor of exchanging several emails with Mike Grehan about some of the theories spreading around the forums and found in many articles at SEO sites. One such theory we discussed was the concept of "themed sites" and how you need a site within a specific theme to stay within that theme in order to rank above your competitors (keeping every other variable as equal). So people were taking the extreme action of buying a new domain name for each theme. Then only having information about that theme on that new domain.

Now Mike is about to release his 3rd edition of his search engine book and he has been researching some of these concepts in great detail. With his permission, I will try to convey what his thoughts are on this topic, but not take away the deepness of the "why" found in his book.

Mike asked Daniel Dulitz from Google this question on 'theming' (is that a new word?). ""Utility" and "depth" really should be measured by a
site's users." What I understand this to mean, based on the examples given by Daniel, is that a site will not hurt in rankings if it contains pages off topic to the real essence of the site.

Example; I have a site on the Smurfs (why did I pick the smurfs, I have no idea) but on that site, I have a page or two on how I block spam with spamassassin, will that hurt my site for ranking well with Smurfs? Not at all, according to Daniel.

Now, will a site with detailed information on spamassassin rank better then the page found at the Smurfs site? Of course, at least that is the goal of the search engines.

So when people discuss themed sites and how they are the only way to rank well, it is believed that some are misunderstanding the concept of what a "theme" is. The search engines think of it as the "utility" and the "depth" of the site (in pages and content).

Disclaimer: I hope I did justice to what Mike has graciously communicated to me. Also, I do not have a page on smurfs, nor do I have a page on spamassasin.
rustybrick is offline   Reply With Quote
Old 08-25-2004   #2
orion
 
orion's Avatar
 
Join Date: Jun 2004
Posts: 1,044
orion is a splendid one to beholdorion is a splendid one to beholdorion is a splendid one to beholdorion is a splendid one to beholdorion is a splendid one to beholdorion is a splendid one to behold
Exclamation

The way I see this, they are right, as long as we don't go overboard with multiple themes in a single page or site.

There is nothing wrong with having a main theme in a page (or links) and secondary themes, as long as they follow a well structured hierarchical arquitecture --from the semantic standpoint, I mean.

The same ordering applies to primary, secondary, terciary keywords (and so on) in a document. In this way we avoid semantic fights (between themes and keyword weights)

Orion

Last edited by orion : 08-25-2004 at 12:18 PM.
orion is offline   Reply With Quote
Old 08-25-2004   #3
orion
 
orion's Avatar
 
Join Date: Jun 2004
Posts: 1,044
orion is a splendid one to beholdorion is a splendid one to beholdorion is a splendid one to beholdorion is a splendid one to beholdorion is a splendid one to beholdorion is a splendid one to behold
The themed sites level concept also plays well with the new ranking algorithm developed by Microsoft on passages or block-level link analysis.

http://forums.searchenginewatch.com/...read.php?t=832

Orion
orion is offline   Reply With Quote
Old 08-25-2004   #4
critter
Underpromise; Overdeliver
 
Join Date: Jun 2004
Location: London, UK
Posts: 286
critter is a jewel in the roughcritter is a jewel in the roughcritter is a jewel in the roughcritter is a jewel in the rough
There are few GOOD SEO books out there.

Is this one worth the $100?

CRITTER
critter is offline   Reply With Quote
Old 08-25-2004   #5
seomike
Md_Rewrite Guru
 
Join Date: Jun 2004
Location: Dallas, Texas but forever a Floridian!
Posts: 627
seomike is a splendid one to beholdseomike is a splendid one to beholdseomike is a splendid one to beholdseomike is a splendid one to beholdseomike is a splendid one to beholdseomike is a splendid one to beholdseomike is a splendid one to behold
Mikes book is great! It really clears up the hub bub you read in the forums.
seomike is offline   Reply With Quote
Old 08-25-2004   #6
rustybrick
 
rustybrick's Avatar
 
Join Date: Jun 2004
Location: New York, USA
Posts: 2,810
rustybrick has much to be proud ofrustybrick has much to be proud ofrustybrick has much to be proud ofrustybrick has much to be proud ofrustybrick has much to be proud ofrustybrick has much to be proud ofrustybrick has much to be proud ofrustybrick has much to be proud of
Agreed, Mike's book goes beyond the "how-to" rank well in search engines and explains the "why am I ranking well or poorly in search engines."

I find the book to be an invaluable resource for myself.

I normally do not say anything in this fashion, but here it goes. I believe its the best and most comprehensive SEM/SEO book available today.

Just to put my moderator hat back on...

This thread is not about Mike's book, its about the themed sites theory. If you want, please start a new thread in some other forum about the book.

Thanks.

Last edited by rustybrick : 08-25-2004 at 04:26 PM. Reason: added part on moderation
rustybrick is offline   Reply With Quote
Old 08-25-2004   #7
seomike
Md_Rewrite Guru
 
Join Date: Jun 2004
Location: Dallas, Texas but forever a Floridian!
Posts: 627
seomike is a splendid one to beholdseomike is a splendid one to beholdseomike is a splendid one to beholdseomike is a splendid one to beholdseomike is a splendid one to beholdseomike is a splendid one to beholdseomike is a splendid one to behold
I've never believed that sites have to stick to theme. I've seen a number of examples where an seo will have a PR generator site on topic A but have a handful of dedicated pages on topic X that links to their client. The example I saw had a topic X page ranking in the SERPS next to the client that it linked to

It would be like having a page about casinos ranking for casinos on a site about starwars.

The only problem with this I see is a possible dilution of theme on any page on topic A linking to the topic X.

When you get right down to it this is the topic that seperates a good seo from an ok seo. It is possible to get a site to rank for multiple terms no matter what the competition is because theme is a page based issue.

The biggest mistake I've seen is making key pages like the index page try and try and rank for all themed terms as if it were a super theme hub.
seomike is offline   Reply With Quote
Old 08-25-2004   #8
rustybrick
 
rustybrick's Avatar
 
Join Date: Jun 2004
Location: New York, USA
Posts: 2,810
rustybrick has much to be proud ofrustybrick has much to be proud ofrustybrick has much to be proud ofrustybrick has much to be proud ofrustybrick has much to be proud ofrustybrick has much to be proud ofrustybrick has much to be proud ofrustybrick has much to be proud of
Quote:
Originally Posted by seomike
The biggest mistake I've seen is making key pages like the index page try and try and rank for all themed terms as if it were a super theme hub.
Maybe not the biggest mistake, but one of the most common mistake by SEO novices.
rustybrick is offline   Reply With Quote
Old 08-25-2004   #9
searchengineblog.com
There's an anti. Let's up it.
 
Join Date: Jun 2004
Location: Above my reputation
Posts: 62
searchengineblog.com has a spectacular aura aboutsearchengineblog.com has a spectacular aura about
For those interested in this topic, Mike covered the themes aspect in my interview back in 2002.

Certainly an interesting topic.

Last edited by searchengineblog.com : 08-25-2004 at 11:53 PM.
searchengineblog.com is offline   Reply With Quote
Old 08-26-2004   #10
Robert_Charlton
Member
 
Join Date: Jun 2004
Location: Oakland, CA
Posts: 743
Robert_Charlton has much to be proud ofRobert_Charlton has much to be proud ofRobert_Charlton has much to be proud ofRobert_Charlton has much to be proud ofRobert_Charlton has much to be proud ofRobert_Charlton has much to be proud ofRobert_Charlton has much to be proud ofRobert_Charlton has much to be proud ofRobert_Charlton has much to be proud of
Quote:
"Utility" and "depth" really should be measured by a site's users."
I'm going to guess that this translates to "relevant inbound links."

And mechanically... narrow site focus also helps internal navigation do double duty. When I build a site that's focussed narrowly enough that words and synonyms for target phrases overlap a lot, I find that the pages start helping each other. Internal navigation links to one page, eg, also help on the page they're from.

For me the jury's still out on whether a link from an on-topic page is worth more than a link from a page that's not on topic, and whether this chain of relevance can pass from page A to B to C.

Not sure whether this is what is currently meant by "theming"... I'm sure it's not what was thought of originally... but it's an interpretation and an approach that I've found helpful.
Robert_Charlton is offline   Reply With Quote
Old 08-26-2004   #11
Wail
Another member
 
Join Date: Jun 2004
Posts: 247
Wail will become famous soon enoughWail will become famous soon enough
Not sites, or pages, but nodes.

It's not so much about themed sites as it is about themed nodes. Google uses pages as nodes and so it's important that your page stays on theme.

MSN released a recent paper on block level nodes. Once search engines operate at that level then the real debate about themes/on-topic will begin.
Wail is offline   Reply With Quote
Old 08-26-2004   #12
orion
 
orion's Avatar
 
Join Date: Jun 2004
Posts: 1,044
orion is a splendid one to beholdorion is a splendid one to beholdorion is a splendid one to beholdorion is a splendid one to beholdorion is a splendid one to beholdorion is a splendid one to behold
Exclamation

Hi, there.

The concept of nodes is not new, comes from graph theory and is quite generic. In some models, the nodes are root urls. In others are just links across a web graph. The links are represented as nodes (coordinated points), the connectivity between nodes is represented as straight lines and the entire network is considered a subgraph of the Web.

Microsoft's block-level model is a model for links placed in specific portions of documents. Those portions are called blocks or passages. In this model themes are represented as nodes and come with two scenarios: page-to-block, block-to-page.

We have discussed the MSN paper at this SEW thread
http://forums.searchenginewatch.com/...read.php?t=832

There is another paper entitled "Block-Based Web Search" which I found a bit more complete and equally interesting. It can be found here http://research.microsoft.com/asia/d...oup/ims/21.pdf

In this paper the MSN team compares four segmentation models with traditional passage retrieval methods. These are:

Web Page Segmentation
Fixed-length Page Segmentation
Document Object Model Segmentation
Vision-based Page Segmentation

Finally, the researchers discuss a combined approach and conclude as follow

"We evaluated the effectiveness of these
page segmentations for block-level query expansion and retrieval,
and verified that page segmentation can significantly improve the
retrieval performance by dealing with the multiple-topic and
mixed-length problems of web pages. Unlike fixed-windows
great importance to plain-text retrieval, such a semantic partition
is more important to the web context. By integrating semantic and
fixed-length properties, we could deal with both problems and
achieved the best performance."

Thus for an SEO, the real challenge is to understand how semantic co-occurrence, keyword weights, themes and relevancy is optimized in blocks (or passages) and at the level of individual documents. We always come back to two things: that IR systems must evaluate information at the global (database) and local (document) levels.

Orion

Last edited by orion : 08-26-2004 at 11:32 PM.
orion is offline   Reply With Quote
Old 08-27-2004   #13
Wail
Another member
 
Join Date: Jun 2004
Posts: 247
Wail will become famous soon enoughWail will become famous soon enough
I agree, it's not new. I do think how the search engines view the web in terms of nodes is pertinent to how they view the web contextually.
Wail is offline   Reply With Quote
Old 08-27-2004   #14
fathom
Member
 
Join Date: Jun 2004
Location: Nova Scotia, Canada
Posts: 475
fathom is a jewel in the roughfathom is a jewel in the roughfathom is a jewel in the rough
Quote:
Originally Posted by rustybrick
Mike asked Daniel Dulitz from Google this question on 'theming' (is that a new word?). ""Utility" and "depth" really should be measured by a
site's users." What I understand this to mean, based on the examples given by Daniel, is that a site will not hurt in rankings if it contains pages off topic to the real essence of the site.

Example; I have a site on the Smurfs (why did I pick the smurfs, I have no idea) but on that site, I have a page or two on how I block spam with spamassassin, will that hurt my site for ranking well with Smurfs? Not at all, according to Daniel.
I think this is being looked at on the wrong side.

Example: what theme is DMOZ?

"All"... why?

Some might say "directory" but then in this instance - logically then only other directories would be included.

To a great extent "scope" defines how strong a single theme but all websites (or most) tend to have multiple themes (sub-themes) and in the hierarchical scope of things the "deeper" a specific theme is in a web site - the weaker it is - but at the same time... the more specific that theme is as well.

Forum are the best example of themes with categories - then forums and sub forums then threads... does a thread here on SEMPO hurt the Google forum - of course not... so why would either hurt a doll site... the other way around the doll site really doesn't help much to promote a SEMPO page... unless it had a major focus toward SEMPO.
fathom is offline   Reply With Quote
Old 08-27-2004   #15
cariboo
Member
 
Join Date: Jun 2004
Location: Paris, France
Posts: 33
cariboo is on a distinguished road
themes in pages or themed sites

Mike Graham pointed out, in his interview in searchengineblog.com, that nobody in the IR research seriously tried to deal with themes for a whole site.

I feel the same.

Search engines deal with pages, and it is more convenient to try to analyze what theme a page is about, than a whole site. Some people already mentioned how difficult it is to associate a site to a "theme". And I doubt it is useful to try to...

For a page, it can be useful (adsense is a good example of a way to use it, and applied semantics technology proved in the past, with Oingo search engine, that it works well for IR as well).
cariboo is offline   Reply With Quote
Old 08-27-2004   #16
orion
 
orion's Avatar
 
Join Date: Jun 2004
Posts: 1,044
orion is a splendid one to beholdorion is a splendid one to beholdorion is a splendid one to beholdorion is a splendid one to beholdorion is a splendid one to beholdorion is a splendid one to behold
Exclamation

I hope this help to clear the above picture.


1. About nodes

Search engines that use the link structure of the web (eg. Google with its original Page Rank and others) use the concept of nodes to represents links (no themes, no content of web pages, just links). The use of nodes is mechanical, ie, to "walk" a subgraph of the Web. This approach is not exactly concerned with the semantic of the web documents or topics. For semantic, other algos are implemented.

The way I see it, analysis of keywords in links and later Google's Topic Sensitive PageRank were attempts to incorporate the semantics of Web pages into their node picture. [I must conceed this is an over simplification of what they actually tried to do.]


2. About MSN's themes

Microsoft's block-level link analysis is novel and different to previous models because uses the concept of nodes to represent segmented themes. Their four different variants of segmentations of Web pages make theme-specific links (therefore localized links; ie., in a given block or passage) highly important to the retrieval process.

In a vector representation, themes can be identified as principal components. From a block-level standpoint -and here is where MSN's model is very relevant- these principal components are even easier to pin-point.


3. About page themes before MSN's paper

Before MSN's paper, themes were treated in the context of hubs and old methods of passage retrieval. Then, SEOs misunderstood the whole concept and came with a theory about themes, which is the one Mike dissected and analyzed.


4. About page themes after MSN's paper

Now MSN came with a better approach for web themes: pages-to-blocks and blocks-to-pages. MSN new approach is more complete and can be applied to PageRank- and Hit-based models. To be honest, it can be applied to any scoring model that deals with localized semantics.



MSN's model is still in the lab and tested under controlled conditions. Time will tell, whether or not it can be diluted (or gamed) by the commercial noise found across the Web. Either way, we are talking about theme theory, before and after MSN's paper.


Orion

Last edited by orion : 08-27-2004 at 11:40 AM.
orion is offline   Reply With Quote
Old 08-27-2004   #17
I, Brian
Whitehat on...Whitehat off...Whitehat on...Whitehat off...
 
Join Date: Jun 2004
Location: Scotland
Posts: 940
I, Brian is a glorious beacon of lightI, Brian is a glorious beacon of lightI, Brian is a glorious beacon of lightI, Brian is a glorious beacon of lightI, Brian is a glorious beacon of light
This theory that's doing the rounds...it wouldn't be ultimately suggesting that good quality on-topic links are actually beneficial to a site, would it? This will shock the SEO community to its core.
I, Brian is offline   Reply With Quote
Old 08-27-2004   #18
orion
 
orion's Avatar
 
Join Date: Jun 2004
Posts: 1,044
orion is a splendid one to beholdorion is a splendid one to beholdorion is a splendid one to beholdorion is a splendid one to beholdorion is a splendid one to beholdorion is a splendid one to behold
Exclamation

Hi, there.

True. Not only high-quality links must be on topic but at which depth. The model also take into consideration where -ie., in which block and block level- both the theme and link appear. This open the door to many applications and other areas in which the model can be incorporated.

For instance, we have discussed block-level analysis using anchors and geo-tags. Imagine incorporating this model to geo-tagged anchor links.

See post #4 of
http://forums.searchenginewatch.com/...read.php?t=636

See post #17 of
http://forums.searchenginewatch.com/...read.php?t=832

Orion
orion is offline   Reply With Quote
Old 08-27-2004   #19
Nacho
 
Nacho's Avatar
 
Join Date: Jun 2004
Location: La Jolla, CA
Posts: 1,382
Nacho is a splendid one to beholdNacho is a splendid one to beholdNacho is a splendid one to beholdNacho is a splendid one to beholdNacho is a splendid one to beholdNacho is a splendid one to beholdNacho is a splendid one to beholdNacho is a splendid one to behold
Using theming for SEO with the Hispanic Market

Actually, I consider theming one of the BEST and most effective ways to target the US Hispanic market when you are reaching the general market on the same website at the same time.

Many marketers and SEOs say, "I don't want to target to Hispanics because I have to translate my entire website into Spanish." -> THIS IS NOT TRUE.

Identify your high traffic and money words and find the correct translation. Notice that I say "correct" because many many times direct translation might be something completely different (for example: PEN = Pluma or boligrafo, however "pluma" means both "feather" and "pen" in English. Are you selling feathers or pens? Get the idea).

Then build a theme section relevant to that Spanish keyword and use that to increase your traffic reaching the Hispanic market. Here is a sample image taken from my presentation at SES San Jose:


This way, you don't have to translate your entire website into Spanish. Plus, these theme sections make ideal landing pages for your PPC campaigns reaching US Hispanics as well ("kill two birds with one stone"). How much you want to follow through all the way down to a complete Spanish checkout will depend on the level of conversions and degree of exits among your users from their paths taken. It might be necessary and it might not be, it's all depending on the market you are targeting (for example, U.S. vs. Latin America) and the level of acculturation of your users.

I hope this helps some of you by using "theming" to target NEW markets, like the U.S. Hispanic market. I wish you success and $$$$ mucho dinero.

Saludos,

Nacho
Nacho is offline   Reply With Quote
Old 08-27-2004   #20
orion
 
orion's Avatar
 
Join Date: Jun 2004
Posts: 1,044
orion is a splendid one to beholdorion is a splendid one to beholdorion is a splendid one to beholdorion is a splendid one to beholdorion is a splendid one to beholdorion is a splendid one to behold
Exclamation

Hola, Nacho.

Adding to Nacho's excellent post, the use of geo-tagged anchors and themes plays well with regionalisms in specific audiences. For example, the Spanish word "cuero" means leather, however we have the following peculiarities.

In Cuba, cuero means drums.
In Puerto Rico, cuero also means prostitute.
In Mexico, cuero is also a nice thing to say to a girl.

So if you say something like, "Que cuero de mujer!" in Mexico sounds like flirting with a girl but in Puerto Rico is more than an insult. The word "coger" in Spain and in the Caribbean (Puerto Rico, Dominican Republic, etc) means "to take" and "to grab" but in Mexico and South America means -let put it in this way- "to do it" and the four-letter word. I apologize for the crude examples. For other examples, check http://www.satyam.com.ar/osho/palabrotas.htm

This reminds me of what Manfred Schroeder wrote in his book "Fractals, Chaos, Power Laws" (page 37), when he discusses Zipf's Law and natural language semantics:

"...the interesting instances are 'incognates' (unrelated words) such as the English word STRICKEN, which means 'to knit' in German, or FALTER (a German butterfly) and LINKS (the German 'left'). And what about such triplets as ART, which is a German word for KIND, which may mean MINOR in German, which in turn is a technical term in the theory of determinants (in either language). Finally, a fivefold string: ROT-RED-TALK-STEATITE-SOAPSTONES. Who can conceive sextuplets?"

The point is that, properly focused and on-topic words are very important, so as your target audience and the used cognate or incognate words. We have seen many instances of using translation search engine tools with off-topic or dumb translations (eg., Google, Babelfish, etc). Properly rotated theme-relevant keywords using a geo-tag mechanism could make a difference.

On and on, use or misuse of themes can attract or chase away audience and money (audiencia y dinero).

Orion
orion is offline   Reply With Quote
Reply


Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)
 
Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

vB code is On
Smilies are On
[IMG] code is On
HTML code is Off