Forum Moderators: Robert Charlton & goodroi
Matt Cutts and Amit Singhal Share Insider Detail on Panda Update
...we used our standard evaluation system that we've developed, where we basically sent out documents to outside testers. Then we asked the raters questions like: "Would you be comfortable giving this site your credit card? Would you be comfortable giving medicine prescribed by this site to your kids?"
There was an engineer who came up with a rigorous set of questions, everything from. "Do you consider this site to be authoritative? Would it be okay if this was in a magazine? Does this site have excessive ads?"
...we actually came up with a classifier to say, okay, IRS or Wikipedia or New York Times is over on this side, and the low-quality sites are over on this side. And you can really see mathematical reasons.
..us lab rats talking amongst ourselves about the lab and the experiment(s) helps them too..;-)
On the topic of "thin" pages -- one of the most useful pages on the internet has NO content: Google search home page...
I often wonder if the Google engineers are laughing their heads off listening to all our theories. :)
I do think Google is giving up a little bottom line here and sacrificing their bottom line for quality.
qualityis what they say it is for ..and what you say it is for ..
[edited by: Leosghost at 1:17 am (utc) on Mar 8, 2011]
I've mentioned Hari Seldon in relation to them more than once in past years here ..someone here in a thread recently again made the connection / spotted the similarities to the philosophy of the "shapers" of Asimov's foundation series again recently ..
I personally think that Google is once again leaving webmasters behind and puzzled.
One of the papers that Biswanath Panda and a number of other Googlers published for Google in 2009, described an experiment that Google performed on their advertising system, seeing if they could learn about the quality of ads and landing pages based upon bounce rates associated with clicks on those ads.
The focus of the paper wasn’t so much upon the effectiveness of the ads in the experiment, but rather about the ability of the machine learning system to work on a very large set of data.
For future work, our short term focus is to extend the functionality of PLANET in various ways to support more learning problems at Google. For example, we intend to support split metrics other than those based on variance. We also intend to investigate how intelligent sampling schemes might be used in conjunction with the scalability offered by PLANET. Other future plans include extending the implementation to handle multi-class classification and incremental learning.
We measure the performance of PLANET on the bounce rate prediction problem [22, 23]. A click on an sponsored search advertisement is called a bounce if the click is immediately followed by the user returning to the search engine.
Ads with high bounce rates are indicative of poor user experience and provide a strong signal of advertisement quality.