Wednesday, May 16, 2012

Using Twitter Data to Improve Search Results

Doesn't it seem pretty obvious that it would be possible to use Twitter mentions to improve search results?

I understand you would have to deal with people trying to "game the system."  Google seems pretty good at doing this.

The nofollow attribute was suggested by Google as a way of marking untrusted links.  I don't believe the application of that is great, but that was Google's plan.  One weakness is the nofollow name is not the same message as untrusted link.  This might not have mattered to Google originally but when you add confusion unnecessarily you open the door to problems.

Google then added to the problems by declaring paid for links should also be nofollow.  Again Google confusing the issue - if they want paid links noted as paid it is fine for them to say that is what they would like.  And if they want to treat paid links the same as untrusted links that is their option.  I think it is a mistake but it is their option.  Telling people they are suppose to mark links as nofollow when they are paid seems confused and lame to me.

Lately I have heard sources quote Google as saying we can't follow Twitter links because they told us not to.  And those are saying that using the untrusted link text Google asked them to was how Twitter told Google not to follow the links.

So anyway here we are today and there are many ways for a search engine to decide some Twitter accounts are trustworthy and that links from those accounts are an indication of the merit of the site linked to.  Why would you not use this information?  Even if Twitter told you the links were untrustworthy (because they were worried about sanctions Google might impose if they linked to sites Google didn't like without making those links nofollow) that doesn't mean you couldn't use the data from Twitter links to improve results.

Maybe in the decision to use the term nofollow and then set standards for what Google would do around this link Google has put themselves in a position where they can't follow nofollow links and do what they said they would do.  I am not sure about this.  But if they can't take advantage of useful data to make search results better due to their previous mistake of calling a untrusted link nofollow they should correct that failure.

If sites want to tell Google not to follow links that is fine (to preserve server bandwidth or just because they don't like Google or whatever).  But, I believe, most nofollow links are now used to

   1) avoid sanctions by Google
   2) keep "link juice" for the monetary benefit of the site that makes the link nofollow (ironically, that makes the non-nofollow links "follow" links essentially for monetary gain, which somewhat subverts Google's desire to remove monetary incentive from placing links)

It seems foolish to me to not use information that could make search results better.  It seems to me valuable information to make search results better is now clouded by nofollow links.  I certainly would be using that information if it were up to me.  And my guess is you can use a large part of it even if the past decisions mean you can't follow the links (this likely would make the silly use of link shorteners more of an issue - but at least the urls that are actually shown could be used to improve search results).

I am pretty sure Google ignores useful measures hidden in nofollow data.  I woudl guess the other search sites are ignoring it too, but I am not sure.

Related: Viewing Unpersonalized Google Search Results - Google Rank Patent for Delegation Authority Factors

No comments: