Welcome to WebmasterWorld Guest from 126.96.36.199
Forum Moderators: open
Ask Jeeves president Steve Berkowitz told the E-Commerce Times that...
Teoma ranks sites not by how many links they have received from other sites in general, but by how many links they have received from sites that are similarly specialized.
Life After Free: Search Engines Seek Cold, Hard Cash [ecommercetimes.com]
Popularity: It understands that a single web page is popular by looking at links to it from other web page.
Context: Separately from this, Google understands what a page is about by looking at the page's content (body copy, title, etc) and by the content of the words in the links and near the links that point at the page.
Searching: So when you search, Google pulls out all the pages that are contextually-relevant for your search, then sorts them by their popularity/PageRank (more or less).
Teoma is the same, for the first part about context. When you search, it pulls all the pages that are contextually relevant, just like Google.
Teoma is different for the popularity part. Only after it has a set of contextual results does it then figure out poularity, in order to rank them. It looks at the links within the set retrieved and only in that set.
In short, Google has a popularity component that is called PageRank that is based on all the links to pages from across the entire set of pages it knows about. It applies context separately to this. Teoma popularity rankings are only generated after a set of documents has been retrieved that are contextually relevant to your search.
What BK said.
This is worth a read DiscoWeb: Applying Link Analysis to Web Search [cs.rutgers.edu].
>Only after it has a set of contextual results does it then figure out poularity
I think they still use basic link popularity in addition to the contextual analysis. The main difference I see between them and Google is that at Google you almost need a site in front of you to pull you in to the top 10, at Teoma you can get a push from behind.
The easiest way to see that, is put in your domain name and see what "related sites" the engine thinks you are related too. In Google, that would probably be the ODP link that pops up. Teoma, would be the "related" links at the top. Same with WiseNut.
*cough* It think I'm going to get the last laugh with this "theme" stuff. ;)
thanks Black Knight, NFFC :)
> The easiest way to see that, is put in your domain name and see what "related sites" the engine thinks you are related too. In Google, that would probably be the ODP link that pops up.
I'd make a couple of distinctions. If you just put in a domain name at Google, you'll see what your home page is related to, but other pages inside your web site may be related to other topics. So say I do this:
There you are, Brett -- right at the top. Now I could use the ODP "category" link as you suggest, but all that's telling me is that Google realizes that this particular URL is also listed in its directory.
Instead, it's really the "Similar pages" option that gets Google to use its algorithm to try and figure out what pages seem to be like a particular page, in terms of content.
Doing that gets this:
and we see that Webmaster World is deemed to be like other sites that cover search engines broadly, especially in terms of webmaster advice. Makes sense.
Now let's say I decide to see what happens for other pages from Webmaster World. I click on the [More results from www.webmasterworld.com] link and get:
Then I choose the "Directories: ODP, Looksmart, About.com: Webmaster World" page's "Similar pages" link:
Now the results are much different than what I got for the Webmaster World home page, and that makes sense. This second page is all about directories, rather than search engines in general.
Interestingly, I'd always understood "similar pages" to be looking at a page, understanding the words on the page and trying to find other pages that seemed like it. Instead, looking at these two results, you get more the impression that Google is depending on link
> Teoma, would be the "related" links at the top. Same with WiseNut.
Agree entirely. And the super easy way to check on what's related to any page at Google is to skip the Similar pages links and just use the related: command, such as:
AltaVista does it:
Haven't seen any power commands like the others, but they might be out there.