| 3:03 pm on Jan 21, 2002 (gmt 0)|
The vote function seems related to the Open Grid Project.
You can find research paper about Website ranking here:
Is this guy already working at Google ? :)
| 4:02 pm on Jan 21, 2002 (gmt 0)|
"Invented a new model, the voting model, for computing quality ranks of Web documents, that improves on the model used by Google search engine.
"So yes it lookes like he did some of it.
| 7:22 pm on Jan 21, 2002 (gmt 0)|
If I remember correctly, GoogleGuy suggested that the Toolbar voting buttons were producing interesting results, and that spam attempts were easy to spot.
My own view is that if Google let it be known that they were using votes for ranking purposes, or if we found that they were spidering sites from Toolbar clicks (and let's face it there's probably no one here who hasn't visited a 'secret' URL to check for this) then its usefulness would very quickly plummet. There are plenty of IPs out there...
ROLAND_F, isn't the Yuntis work about 'voting' in terms of links rather than buttons?
As far as I can tell, Lifantsev's work is to do with extending the citation ranking approach further in the direction of hub and authority, as well as some kind of hybrid.
| 1:35 am on Jan 22, 2002 (gmt 0)|
I just enjoy using the "vote against this page" button on spammers homepages, and sites that send me spam. Very satisfiying.
Scared to use the "vote for this site" on my own site...
| 2:37 am on Jan 22, 2002 (gmt 0)|
I wonder just how much google is using the voting data, and how much they are sorting it.
I wonder if they'd do something like the following:
For all voters who never cast positive votes, lower the value of those votes because the user is overly critical.
For all voters who always cast positive votes, lower the value because the user is overly positive.
For all voters who have voted less than 5 times, exclude those votes until votecount > 5 because those user's arent active enough(to gather usage data).
Or google could group votings by site,
whereby every voter has a certain initial "sway", and so thier sway is. distributed amongst all the sites they have voted on.
this way no one vote can overly skew results,
ie: similar to the pagerank algo but inbound links are votes & outbound is the voters total sway or similar.
Or combine the two.
Or i'm wrong
Or i'm close to right
Or i have an obsession with the word or,
Or i forgot to capitilize my I's
Or i cant think of another or <grin>.
| 1:23 pm on Jan 22, 2002 (gmt 0)|
At last, I have a really good use for the smiley data.
[big]"People who like this page also like these pages..."[/big]
I guess it would be computationally expensive to run, but it would be very useful and to cheat you'd need to have a sizeable % of the smiley clicks on popular sites, not just your own.
So to come up in the drop-down for yahoo.com you'd need a whole load of :):):):):):)s
| 3:02 pm on Jan 22, 2002 (gmt 0)|
It would be a small improvement over what Alexa do.
They could also do something like Daypop.com .
I wonder why can you only vote for or against a page.
Most ranking sheme we can see on the web have a larger scale
such as the HotorNot website.
It would be nice to see what theory is behind their voting system.
| 4:22 pm on Jan 22, 2002 (gmt 0)|
#3, #5.3,: [www10.org...]
|...that defeats every other in pairwise simple majority voting, then that this element should be ranked first |
[theory.stanford.edu...] (<-notice the homepage its under)
| 4:36 pm on Jan 22, 2002 (gmt 0)|
Thank you very much Brett, very nice paper!