Search Engine Optimization October 22nd, 2008
It’s tiresome when in absence of a proper academic curriculum we learn the SEO ropes from rehashed paraphrased stuff from ‘been-there-done–it’ SEO’s. While there is nothing wrong with that, once in a while you crave for the pure pristine stuff. You know – the stuff right from the source. And what better source to go to for pure SEO theory then Google itself. Ah the feel of that pure dark gooey chocolate in your mouth. Yum Yum!
Today – I researched this patent from our good friends at Google.
Patent number: 6526440
Ranking Search Result By Re-Ranking the Results Based on Inter-Connectivity
Filing date: Jan 30, 2001
Inventor: Krishna Bharat
Assignees: Google, In
We have known for some time that Google, to deliver better overall search results and put core factors of ranking out of SEO’s manipulation chooses some websites as ‘favored’. They may be human reviewed and then picked or as this patent suggests, computed internally by the algorithm itself. It shines a floodlight on what has so far been very ambiguous; aspects related to how a website is considered ‘favored’ and how that status influences the SERP’s.
On a query request a set of websites are pulled from the Google database and sorted on the basis of a relevancy score assigned to each website. Assigning a relevancy score is the first step in this process. Since Google is super at refining the process to get highly relevant results there is a further sorting of the websites to achieve that.
This is accomplished by finding at least two documents or websites to which there is a consistent pattern of linking from other relevant websites within a set. This relationship is quantified and called Local Score Value (LSV). We are not told the details of the relationship factors considered for calculating the LSV. This provides the search engine with a benchmark to evaluate other relevant websites further.
This is really a dual layer purifying process that burns away all the dross. It’s also made very clear that only one website from the same host (IP) or from affiliated hosts become eligible and the one with the lowest relevance score is removed. (ever wonder why you have 2 websites that are similar and only 1 of them ranks? This is why!)
We are also told how this LSV is computed.
Using this benchmark value arrived at, Google assign to the second set of documents an LSV score.
Here is the geek mix i = 1 k OldScore (BackSet ( i ) ) m
Google calls it “old score” = old score is the relevance score of that particular document
‘Backset’= Backset has a few separate calculations that are brought together to arrive at ‘Backset’
In English? LSV means to put it in earthly tongue, the original relevance scores (old score) are recalculated with LSV. We are also told that other factors to arrive at the eventual score and determine the ranking are size of the set of documents and the maximum relevance score of all the documents in a set.
· If you have been wondering why Wikipedia dominates the higher echelons of search result the answer may very well be in this post. Local Score Value could be calculated from the linking and other patterns displayed by that article in Wikipedia.
· Bad neighborhood websites are a big ‘no-no’. What is required is more thematic and topical link building strategy.
· Content was king but it has been dethroned (somewhat) now as context rules. Now keyword is passé and it’s now moreso the idea or a theme. Your theme (market +keyword) is determined by who links to you.
I am sure there is much more to take out of this patent. In my next post, I will be discussing some very interesting research coming in from University of Toronto that I believe is going to further improve the quality of search and make our lives more, umm, interesting. But then at least we know what’s coming.
Time for me to go gather some more ‘Gooey Chocolate’!
How Google may be refining Search by choosing favored WebsitesRead time: 3 minutes