Google: LocalRank

[Brian Turner]

Here's another important feature that is also expected to be applied to Google results:

LocalRank patent

Also going under the name of LocalScore, here's a quote from the paper:

What is claimed:

1. A method of identifying documents relevant to a search query, comprising:

obtaining an initial set of relevant documents from a corpus;

ranking the initial set of documents to obtain a relevance score for each document in the initial set of documents;

calculating a local score value for at least two of the documents in the initial set, the local score value quantifying an amount that the at least two documents are referenced by other documents in the initial set of documents; and

refining the relevance scores for the documents in the initial set based on the local score values.

2. The method of claim 1, wherein calculating the local score value for a particular one of the relevant documents further includes:

forming a sub-set of documents from the initial set of documents as the sub-set of documents that contain a hyperlink to the particular one of the relevant document, and

removing documents from the sub-set that are from the same host or from an affiliated host as the particular one of the relevant documents.

3. The method of claim 2, further comprising:

removing, for each pair of documents in the sub-set that are from the same host or an affiliated host, one of the documents in the pair that has a lower relevance score than the other of the documents in the pair.

4. The method of claim 1, wherein the local score values are based on the relevance scores.

5. The method of claim 3, wherein a predefined number of the documents in the sub-set are used to calculate the local score value.

6. The method of claim 3, wherein the local score value is calculated for the particular one of the relevant documents as: ##EQU2##

where OldScore(x) refers to the relevance score value for the particular document, BackSet refers to the sub-set of documents, the sum is taken over the first k documents in BackSet, where k is a predefined number, and m is a predetermined constant.

7. The method of claim 6, wherein refining the relevance scores is based on taking a product based on the local score values and the relevance score values.

8. The method of claim 6, wherein refining the relevance score values for the documents further includes:

recalculating the relevance score values for the documents as

NewScore(x)=(a+LocalScore(x)/MaxLS)(b+OldScore(x)/MaxOS)

where NewScore is the recalculated relevance score value of the particular document, a and b are predetermined constants, MaxLS is equal to the maximum of the calculated local score values, MaxOS is equal to the maximum of the calculated relevance score values, and LocalScore(x) refers to the local score value of the particular document.

9. The method of claim 8, further including:

setting MaxLS to a predetermined threshold value when MaxLS is below the threshold value.

10. The method of claim 1, wherein obtaining the initial set of relevant documents from the corpus includes obtaining the initial set based on a matching of terms in the search query to the corpus.

11. A method of responding to a search query from a user, the method comprising:

receiving the search query from the user;

generating a list of relevant documents based on search terms of the query, each document in the list being associated with a relevance score corresponding to a relevance of the document;

calculating a local score for documents in the list of relevant documents, the local score quantifying an amount of inter-connectivity between documents in the list of relevant documents;

refining the relevance score based on the calculated local scores; and

returning a list of relevant documents to the user, the list being sorted based on the refined relevance scores.

12. A system comprising:

a server connected to a network, the server receiving search queries from users via the network, the server including:

at least one processor;

a database of a corpus; and

a memory operatively coupled to the processor, the memory storing program instructions that when executed by the processor, cause the processor to: generate an initial list of relevant documents from the corpus based on a matching of terms in the search query to the corpus, rank the generated list of documents to obtain a relevance score value for each document in the generated list of documents, calculate a local score value for the documents in the generated list, the local score value quantifying an amount that the documents are referenced by other documents in the generated list of documents, and refine the relevance score values for the documents in the generated list based on the local score values.

13. A system for identifying documents relevant to a search query comprising:
means for obtaining an initial set of relevant documents from a corpus based on a matching of terms in the search query to the corpus;

means for determining a relevance score for each document in the initial set of documents;

means for determining a local score value for the documents in the initial set, the local score value quantifying an amount that the documents are referenced by other documents in the initial set of documents; and

means for refining the relevance scores for the documents in the initial set based on the local score values.

14. A computer-readable medium storing instructions for causing at least one processor to perform a method that identifies documents relevant to a search query, the method comprising:

identifying a set of relevant documents from a corpus based on the search query;

ranking the set of documents to obtain a relevance score for each document in the set of documents;

calculating a local score value for the documents in the set, the local score value quantifying an amount that the documents are referenced by other documents in the set of documents; and

refining the relevance scores for the documents in the set based on the local score values.

What does all this mean?

It means that IPs that share all but the last set of numbers are essentially going to be penalised if this algo is applied by any Search Engine (note the author is Google staff). So stuffing lots of backlinking sites on just one server IP would have little SEO benefit by this methodology.

[sarahk]

Originally Posted by Brian Turner

It means that IPs that share all but the last set of numbers are essentially going to be penalised ...

While I can see that IP is a good way of identifying the guys who buy space and load it up with spam consider my situation.

I have a collection of customers who refer others in their industry to me. I recommend hosting company X because they're good. These customers aren't good at linking strategies but in their articles they refer to other companies who provide good related services - but woops, they're on the same server. What might work in the USA won't necessarily work in NZ where we have 3 classes of website owners

• Conservative beginners - all hosted by 3 companies
• Outsourcing beginners - hosted by their web designer, so large range of IP
• Do-it-yourself-ers - back to limited hosting as the range of locally hosted, reliable, CGI/PHP enabled companies shrinks

Now the Google techs are pretty smart so I'm sure they've considered this.

Sarah

[Brian Turner]

Hi Sarah, and welcome to the Platinax business & internet forums.

Indeed, and I have a funny feeling that if Google has ever implemented LocalRank, then it has done so only in conjunction with other automated ranking criteria which might try and take into account such issues.

However, for those who are really focussed on link building it does raise an important point that if you are going to use sites for linking purposes, then you could likely get the most mileage and effect out of them by hosting across multiple C classes.

The US webhosting market is also pretty cheap, and there's a lot of opportunity there for spreading sites and costs around.

[PriceComparison]

I do not agree that google is really that stupid of using IP address to base their website geographical area. I believe 70% of the websites in the world are hosted at datacenters / colocations. Becuase of this their Geo IP location is not accurate!

Andrew

[Brian Turner]

Indeed, there are limitations in how IPs can be used - that's why Google will rank a .co.uk hosted in the USA in Google.co.uk UK only searches, even though they are obviously not hosted in the UK (although Ask does the reverse and only lists .uk domains on Ask.co.uk UK searches).

However, in terms of processing the link popularity between websites, Google does have tools for develauing cross-links between sites for link popularity purposes. There was an old SEO practice of hosting lots of sites on the same server, and simply cross-linking them to gain link popularity. LocalRank is just one potential tool Google can use to devalue such artificial cross-linking.

Hope that makes things a lot clearer.