|On theme page on off theme site|
Is this a themed link?
| 6:41 pm on Sep 22, 2004 (gmt 0)|
I'm aware that a link from a website with a completely unrelated theme is probably of little value - some people even speculate that it could actually be damaging.
But what about the following ...
My widget website gets a link on a page like this ..
In other words, the site itself is nothing to do with widgets, but they happen to have a links page with information about widgets, with links to other widget websites.
So my site has a link on a page that has the filename 'widgets.html', the page title 'widget websites', widgets in the page headings and text etc plus links to other widget sites.
Would this be classed as an on theme link - or does the whole domain need to be on theme?
| 12:13 am on Sep 23, 2004 (gmt 0)|
If you are chasing PR, that is a mathematical calculation not a question of relevancy.
If you are seeking help in ranking from anchors then relevancy is vital.
We have a load of test sites and now accept that the page on which the link is placed should be relevant, not necessarily the site.
| 6:42 am on Sep 23, 2004 (gmt 0)|
Thanks for that Granpops
Yes, I've been thinking that would be the case - consider getting a link from dmoz for example.
Presumably, dmoz has the theme 'web directory' - but getting a link from dmoz.com/directory/yoursubject.html does not give your site the theme of being a web directory!
| 7:31 am on Sep 23, 2004 (gmt 0)|
Yes i also have this feeling the google considers the page not the entire site and a link for travel site from a shopping site with travel related page is absolutely on theme as far as google is considered. But there has been a talk of links passing their effect on furhter pages so the other pages of the site where this page will definitely have links can change this concept altogether
| 8:23 am on Sep 23, 2004 (gmt 0)|
|I'm aware that a link from a website with a completely unrelated theme is probably of little value - some people even speculate that it could actually be damaging. |
TieBreaker: who told you such a thing as fact?
Because it is not even close. There have been no studies proving that themed links are *any* more effective than a 100% non-relevant link, or that a non-relevant link can damage you at all.
Anyone that wants to debate that point, I ask you to show it right here in this thread with provable data to back it up.
To date there is no evidence at all of theming.
I realize some of you think the search engines are artificial intelligence of some sort or another, but sorry to say, they are not. Not yet anyway. They are simply programs whose core functions are to spider, index, and search with some extra filters added to look for certain things that are common traits.
Those who wish to use CIRCA as a reference for possibility, that is all it is at this point, a future possibility with nothing to show it is advanced enough yet to be implemented.
As for PageRank, as Granpops pointed out that is simply a mathematical calculation based on links and their originating PR.
| 10:20 am on Sep 23, 2004 (gmt 0)|
People keep coming up with ideas that would make sense for a search engine to incorporate. The ideas get discussed and, before you know it, some people begin to think of them as fact. The subject of this thread is one of those ideas.
What people usually fail to realise is that search engines cannot simply incorporate 'good' ideas how ever much they'd like to. There are technical reasons why they can't do it. For instance, Google is unable to auto-detect hidden text and auto-redirects, and those things have been around since way before Google.
Recognising relevant links is very difficult because they would need to recognise what is actually relevant. For instance, is a car hire page relevant to a hotel page? Yes it is, but could a program determine that. CIRCA can usually tell you the topic of each page at the ends of a link, but it doesn't tell you if the pages are relevant to each other, unless they happen to be on-topic with each other. But a car hire page isn't on-topic with a hotel page. A program would decide that a page about hotel equipment is relevant to a page about a hotel, but it isn't.
Ascertaining relevancy is no small feat, and it's something that the engines are, as yet, incapable of doing at the time when it's needed - search time.
There's something else to consider. Search engines need to keep their response times down. At the outset, Google were so keen to keep it down that they specifically stored certain things in memory, because disk seeks took 10 thousandths of a second at the time. Now consider what would happen at search time if link relevancy were incorporated. Google would compile a set of the about 40,000 results for a search query (it's their original figure). Those 40,000 pages would have goodness knows how many links pointing to them - some would have thousands, and the page at the end of each link would need to be checked for relevancy. That's a hell of a lot of checking to be done and a hell of a lot of computing time, which the engines don't have at search time.
You could say that each link's relevancy could be ascertained when parsing each page, but relevant to what? There are many topics that most pages are relevant to, and not just the page's main topic. So each link would need to be identified as being relevant to multiple topics, and that information would need to be stored. I could go into much greater detail, and provide more reasons, about why it isn't feasible to ascertain relevany at parse time, but I'll leave it by saying that it's just toooooooo big a thing for the engines to do.
It is very easy to come up with good ideas and assume that the engines must do them because they make sense, but it's a mistake.
| 2:14 pm on Sep 23, 2004 (gmt 0)|
Not all links are created equal. Maybe for numerical PR and raw anchor text points, but if Google can judge relevancy from links two hops back, we better not think there isn't more to it.
Besides, how about the Topic-Sensitive PR papers from early 2003, and the US patent granted early this year? Are those nothing more than an exercise in futility? Is Krishna Bharat just blowing wind?
| 3:35 pm on Sep 23, 2004 (gmt 0)|
|Is Krishna Bharat just blowing wind? |
I can't smell anything here ;)
As far as I know, there's no evidence that TSPR is being used. Apart from that, TSPR is about topics and not about relevancy. They are quite different.
| 8:08 am on Sep 24, 2004 (gmt 0)|
Thanks for all those interesting responses - the more I think about the immense task involved in calculating relevancy, I'm inclined to agree that theming is not something to spend much time worrying about.
However, it is worth remembering that even if all this stuff about themed links being worth more is rubbish, links from on topic sites are still worth going the extra mile for - purely because your visitors will be more targetted.
That is what links are supposed to be about - people visiting your site.
I don't think I mentioned anywhere that someone had stated this to me as fact - I said that some people 'speculate' that an off theme link could be damaging.
Personally, I've never really been sure either way - I just thought that I'd ask peoples opinions whether they would consider the example I quoted above as an off theme link or not.
| 3:43 am on Sep 25, 2004 (gmt 0)|
Hey Phil, LTNS!
|As far as I know, there's no evidence that TSPR is being used. Apart from that, TSPR is about topics and not about relevancy. They are quite different. |
Relevancy isn't evidenced - there can be a page that's just scraped junk or that reads like it's been run through a jargon generator, but has the proper ingredients in place and it can rank.
It sure doesn't look like TSPR is being used, especially looking at the weight anchor text still carries and checking backlinks on some sites. But the sequencing of papers and patents out there lends suspicion that they've got something on that order in their minds. And then there's stemming and semantics. Those are too prominent in conversation to ignore. Besides, what we see today isn't necessarily what we'll see tomorrow.
Somehow all roads seem to lead back to Jon Kleinberg, including the patent issued this year
That's what got me very nervous (as well as fascinated), especially with some sites disappearing into oblivion during and soon after the Florida debacle. Even whole categories of sites getting hit - there were lengthy discussions on a couple of specifics that were locality oriented. I personally saw one site come back only after relevant outbound and inbound linking was added (locality related, BTW), and two that have not and don't seem like they will. Not that they don't have links and PR - but they do not have any inbound topically related links. Just anecdotal observations, but it leads me to think that we may have to start thinking further than we have in the past.
Not that PR isn't PR, just a number, and that anchor text isn't anchor text, but that at some time an added component of on-topic links may needed. The time may come that it's a necessity to have the *right* links on top of that.
Incidentally, this is a good OLD discussion, back when the landscape was totally different:
What's interesting is Yahoo acquiring all those other search companies. Looking up a hobby site of mine at Yahoo that's on my ISP space, with all the many, many "sites" on the same physical ISP's domain, clicking on "other sites from this domain" at Yahoo brings up only those that are on the same topic as mine, either by theme or by including the keywords on their pages.
Especially with the coming emphasis on local search, I suspect they're all headed toward refining relevancy and/or topical standards. I would like to hear a good, thorough explanation of the difference between the two - topic and relevancy; it's a very thought-provoking concept.
|You could say that each link's relevancy could be ascertained when parsing each page, but relevant to what? There are many topics that most pages are relevant to, and not just the page's main topic. So each link would need to be identified as being relevant to multiple topics, and that information would need to be stored. I could go into much greater detail, and provide more reasons, about why it isn't feasible to ascertain relevany at parse time, but I'll leave it by saying that it's just toooooooo big a thing for the engines to do. |
How about another step thrown in somewhere, maybe based on linkage, that is done at other than query time? Isn't that even a remote possibility?
| 10:16 am on Sep 25, 2004 (gmt 0)|
The engines are much further down the road of semantics than most of us probably realise. Fast had a small semantics-based search engine running quite a while ago.
In a recent discussion, I was effectively accused of being short-sighted because I am looking for what works now, but the engines are continuing to evolve, and the algos continue to change. My reply was that I *am* short-sighted because I want top rankings now - not sometime in the future when the engines have caught up to my optimisation. The engines may never catch up - they may shoot off down a difference evolutionary branch than I'd imagined and catered for. So, if off-topic links work today, they'll do for me.
I always recommend doing what it takes to get top rankings in the current conditions. The main thing that it means is to get links, links, and more links, all with targeted link text. As a 'just in case' measure, make it a point to get some of the links from on-topic pages. They may or may not be essential in the future, but it doesn't hurt to take the precaution now.