Google's "Edit Search results" experiment

http://justinhileman.info/blog/2008/07/googles-edit-search-results-experiment

by msg • 18 years ago

15 10 18 years ago

bprater• 18 years ago

I'd say this is a pretty big deal if they push it into normal SEO results. I'd welcome it. I'm tired of spammy grayhat results I get now. Just clicking [X] would feel good.

I think Google is finally realizing that the algorithms still have a way to go and that the massive amount of good stuff in human's heads is quite relevant.

fallentimes• 18 years ago

A whole new game would begin of "X'ing" competitors search results. They need to tread very very lightly and put many checks and balances in place.

josefresco2• 18 years ago

Yeah see Wikia for an example of this. Monitor any semi-high volume keyword and you'll witness a parade of buries/promotions until 1 side finally gives up.

jnovek• 18 years ago

The newest "AI on the web" strategy seems to center around this stuff -- computers do 95% of the work and rely on actual people when the law of diminishing returns kicks in.

We have so many warm bodies on the internet with this marvelous (and mysterious) sense of intuition that we can't seem to make the computer machines replicate. Why not?

See also: the super cool reCAPTCHA project. Or the FoldIt! protein folding game.

tjpick• 18 years ago

we were doing this stuff for a couple of years at a search company I worked for.

msg• 18 years ago

If you're allowed to talk about it, the data you get back is sparse, right? So how useful does the user feedback end up being? It seems like one of those areas where owning search share actually means something.

DaniFong• 18 years ago

There's a type of machine learning called semi-supervised learning that can make use of sparse datasets like you suggest.

PStamatiou• 18 years ago

as a wise man once said "everything that can be socially powered, will be"

josefresco2• 18 years ago

Isn't monitoring where people actually surf more accurate than allowing those same people to 'vote' on a search result's relevance?

I would think many people would vote one way, and surf another.

philh• 18 years ago

When I'm searching for a solution to a problem, I often have to click through to several results before finding it (assuming I find it at all). In that case, clicking through is not an endorsement; clicking through and then not clicking anywhere else might be, but it might mean I've given up, or it might mean I opened several results at once and found the solution in any one of them.

It might be more accurate in other search domains though.

Google's "Edit Search results" experiment

Related Stories