3 ms·
The first question that came to my mind was: How do they know what is an ad and what is just a heavily graphical site that is exactly what the user was looking
by Silhouette 15y ago
The first question that came to my mind was: How do they know what is an ad and what is just a heavily graphical site that is exactly what the user was looking for?
If they're virtually rendering the page and looking at how much page area is covered by content from known ad networks (including Google Ads, I trust) then that's one thing. If they're looking for lots of nice, Google-friendly text and just discarding images, Flash, etc, then that's another thing entirely.
- avsaro 15y agoI do not think it is relevant to separating ads from useful content. They are probably just calculating place of content which is relevant to search keywords in the screen. This would works for images and other graphical content too.
- MatthewPhillips 15y agoThis is a solved-problem, every ad-blocking add-on already does this. I don't think Google of all companies is returning a lot of false positives in their algorithm.
- shabble 15y agoI wonder how much the in-house expertise of developing Chrome is contributing to the GoogleBot analysis stuff. Having people who aren't afraid to go knee deep in WebKit and produce some nice fast library for partially evaluating layouts, etc, is probably quite a nice thing. As one of the sibling comments here mentions, ad detection is a pretty well-solved problem by now, through size, position and origin matches. The interesting part here I think is how they're actually doing the 'virtual rendering' (and presumably, reflowing it for each of their mentioned typical browser resolutions) to check what's actually 'above the fold'.
- joshu 15y agoi suspect that the cpu time to even partially render a page via webkit is really expensive at scale. that said, i recall there was a headless chrome that someone built - anyone remember where that is?
- joshu 15y agoanswered my own question: http://www.phantomjs.org/ http://www.phantomjs.org/
- Chirono 15y ago| I wonder how much the in-house expertise of developing Chrome is contributing to the GoogleBot analysis stuff If anything, I expect the contribution worked the other way, with the rendering engine being developed purely for search, and then fleshed out into a full browser. There was an interesting post about this a few months ago: http://news.ycombinator.com/item?id=3182579