How does google determine if sites are related?

I keep on hearing how link from a related site (same theme) counts more ...so does anybody have any idea how exactly does google determine how related 2 sites are?

 

Related is probably the wrong word to use, as Google uses this word in its algorithm to mean pages that are related to each other. When Google uses the word 'related' it means mostly pages that are in the same neighbourhood as yours. (the neighbourhood is built up by working out who you link to, and who you links to link to)

I am guessing that in your case though you are asking about contextually related site links being better. There really is no reason why you should be linking from non related sites anyhow, apart from to gain page rank and a false sense of importance for your site (I am not teling you not to do it, I am speaking from a search engine point of view). What earthly reason would there be for a painter and decorator in Berlin to trade links with a butcher in Preston? Other than to manipulate the SERP's that is.

Words are related, and their relationship between words is know as semantics, and it is via this technology that a search engine would decide what is related.

By looking at pages, and comparing the content, an algorithm can make a 'related relevance' score independent of any outside factors, based purely on comparing the content semantically. Add in then the links outgoing, and the links incoming, and you can have a truly great set of SERP's.

E.G. A page well written with quality content about sheep farming in Wales would have all the right words on it that relate to Wales, Sheep, & Farming, so semantically it is about at least three things. It might link out and back to the farmers union, co-operative sheep marketing sites etc. It might link out to animal breed sites, and be linked to by many many other sites about the subject of sheep breeding and farming.

Safe to say that in the above example a semantic algo would have no difficulty in knowing exactly what that site is about, and it would score highly as it is focused about a range of subjects that are semantically related

Sheep
Farming
Animals
Breed of sheep
Agricultural unions
and so on. These words would all turn up as semantically related to each other, and would be likely to appear on the pages of the sites linking to this page (site) in anchor text, and across the range of related pages link wise.

Hopefully that answers the question you asked.

 

As for the related issue with regards to in/outbound links, you can give this tool a try http://www.touchgraph.com/TGGoogleBrowser.html

Very interesting to see who is in your neighborhood and who is in your competitors....

 

yes, i was wondering about how google would determine if sites(pages) that link to you are contextually related to your site, and more specifically how exactly would semantic algorithm work to figure that out..would it be # and/or % of same keywords (and variations) on both sites, and so on......

and the reason why i am asking this is:
i am using linkmarket to exchange links, and everyone there puts reciprocal links deep inside their link/directory pages or on pages with 100+ links. So nobody could possibly expect any real traffic from those links, and the only reason for doing this exchange is hoping that search engines will pick up those links to your site and increase value of your site.
To do this, all you would need is just to give people url and anchor text that you want them to link to you with....BUT, i am getting link requests where they want me to link to their pages using long descriptions after anchor text as well, stuffed with keywords, like 30,40 words even. so i figure they are doing this, hoping to make my link page seem contextually related to their site.

 

Some sites cover a whole range of things (like my BRE group site) so how can Google determine whats relevant and what isn't. Its fine using a easy "sheep farmer" example but there are lots more examples where its more difficult to pigeon hole the site into a certain category. Also things like media and general information sites might cover a whole range of subjects.

I don't see how Google can, with any certainty, decide what is a relevant or themed link and what is not.

 

It is believed by many that G is in fact using Latent Semantic Analysis (LSA) also referred to as Latent Semantic Indexing (LSI) to determine relationships.

There are a number of threads on this forum on the topic and a research paper here: http://www-psych.nmsu.edu/~pfoltz/reprints/BRMIC96.html

/*tom*/

 

And that is why you or I are not working for them

You have used the word 'site' when you should be talking about 'page' for a start. OK Search Engines might theme a site, but as you rightly point out, how can they pigeon hole a site about many things. The simple answer is that they probably don't, they are proposing trust rank for a start. Trust rank is where sites are seeded manually. A quality directory for example could be seeded and as such a link from it will count highly as an on topic link because it is a 'trusted' site that only links on a themed basis.

How many sites are there that can not be themed semntically? I would say not many, if any. A site that is about lots of things is more difficult to theme, but then again, the incoming links are likely to help enourmously to theme sections of the site. the site could be about business, there may be sections on sales and selling, funding, budgeting, software etc, but these are all interelated anyhow.

 

I think their support checks the ad or the google bot might have some options to check it.

 

Sorry, care to explain the post

 

Of course, we can only speculate on whether or not G is using the technology at all. Recently, I became a believer though.

About 3 months ago, I added the phrase 'classroom response systems' to an existing page. I can't say I even optimized for it, as I added only 3 mentions and that was in the body only.

About a month ago, I stumbled across the fact that I ranked #60 or so for 'student response systems' (pool of 12.7 mil) and was quite surprised as the word 'student' did not appear on the page at all (and definitely no anchor).

I am left to suspect that LSA (or similar technology) is responsible for the correlation between classroom and student. Of course, this is an on-page example. But, if LSA is in use, I don't think it's a stretch to apply it to the process of determining link relevance.

That of course assumes that G is in fact making link relevance determinations, and who of us knows for sure? Wish I did though

/*tom*/

 

Maybe he just likes the butcher's site and thinks its cool. Maybe he likes to decorate houses with pictures of butchers shops? OK, so it's not very likely, but really, anything could be related to anything if you are creative and unique. And isn't Google said to reward unique content?

 

Your right, on the ODD occasion there might be a reason for linking to a site that at face value does not appear related. The problem would kick in when the BULK of your links are unrelated. With the greatest of respect, if your site is about a set subject and you are linking to all sorts then what is the value of your link?

Google doesn't reward unique content, it frowns on duplicated content, there is a difference. If you have a site about one subject, and you are linking to other sites that are related to your subject , then your site is likely to be seen as a damn good resource on that subject.

 

The butcher could have had his shop painted and decorated by the Berliner?

My point is just that the whole relevancy thing gets taken a little too far sometimes. I might have a site where friends and I talk about travelling and I mention how one night on a trip we ordered a dominos pizza that was really good from "this pizza place" (where the text in the quotes would be a link to the website). Does that make my link irrelevant? No; just because my site isn't identically themed doesn't mean a link from that site should necessarily be considered less important.

 

Totally agree kkibak. thats the point i was making (badly) earlier in the thread.

I really don't think that Google does or can, with any certainty spot related sites that are linked. The only reason, and its a good reason, for only linking to relevant sites is so that visitors see your site as a great source of information for your subject matter. Nothing to do with SERPS.

 

KKibak, you totally missed what I was saying I feel I just didn't explain it well enough maybe. I am talking about links in general, sure there are cases where there might be a reason to link oddly (or what appears to be oddly), but they are by and large the exception, not the rule.

Quality sites will naturally link to quality sites in the same field. It is natural, it can't be helped. This site we are currently on links to many Industry sites. These links are normally in context where they page linked to is directly relevant to the content on the page linking out. Now if on that same page you linked to your Pizza deliver guy, which link do you think is likely to be more relevant? Keep in mind that we are by and large not talking about things as they stand, I never said that a related link is worth more, I was simply answering a question on how this can be achieved, and how it works.

BobL re: I really don't think that Google does or can, with any certainty spot related sites that are linked.
I suggest you read this, and have a look at some of the patents that the search engines have registered. I am not being awkward here, but believe me when I say that Google CAN spot sites that are semantically linked. Whether or not it is doing this at the moment is another matter for discussion, but the technology and ability is not at question. Why did Google spend all those millions to buy Applied Semantics if not for their technology to do just this?

 

that's why GG is still the best for me. I get the best searches from it.

 

Geez, Chris, this is the 3rd post of you saying exactly the same thing, no wonder you are red at 50 posts...

Anyway, Google is weird, when I do a search for similar sites of my emoticons site, I get this one...

www.digitalpoint.com/tools/counter/

And I don't even use DP's counter! Even if I did, how would it be similar to an emoticons site?

 

Google purchased Applied Semantics to obtain Applied Semantics AdSense product. Up until then Google's 'automated content-targeted ads' as they called them worked on url/web log statistics and user trends and patterns. Applied Semantics AdSense product was far superior in that it used a proprietary (and patented) technology based on an ontology consisting of millions of words, meanings, and their conceptual relationships to other meanings which they called CIRCA (Conceptual Information Retrieval and Communication Architecture).

This is not the same as Latent Semantic Indexing which as it implies is to do with indexing. LSI basically records which keywords a page contains then examines all known pages as a whole to see which other pages contain some of those words. LSI considers pages that have many words in common to be semantically close and pages with few words in common to be semantically distant. Two pages may be semantically very close even if they do not share a keyword.

- Michael