Google has just unbelievable computing power sitting out there on the public internet. But they are also a huge semantic processor. In fact, the search engine companies sniped most of the employees of a recently defunct speech recognition company I knew. The former head of Speech Technology at Microsoft, Kai Fu Lee, is now the head of Google China. Nice guy.
So today, as always, under the aegis of research, it was all about adSense. The thing that is reading this post now, and trying to figure out which ad to put up above. Its totally fun to write about. Because all you have to say is "Wocka wocka wocka" and watch it go into a tailspin. Are fozzy bear and kermit really twins? They look like twins. Sorry I will have to deal with that question later. Right now, I am writing text that will be processed and the adwords above, will directly relate to it!
Its absolutely fascinating / just peek up under the hood. This is the best I could gather. There are two levels of parsing. Site. And Last Post. The Last Post content parser looks at this post. Tries to figure what the general drift could be. Then fits it into ad categories probably by markov chain, until they get a supralexical category match. They will return maybe six categories for this content here, then they will drift around through the weighted scores matrix of that category search and slam the near-match. Into place. They probably placehold that, and then on the next pass, they come back with a re-weight. They do this all the time, like, every seven minutes or so. Ad text always comes from the advertisers. But the categories come from Google. From wikipedia:
The underlying technology behind AdSense was derived originally from WordNet and Simpli, a company started by the founder of Wordnet — George A. Miller — and a number of professors and graduate students from Brown University, including James A. Anderson, Jeff Stibel and Steve Reiss. A variation of this technology utilizing Wordnet was developed by Oingo, a small search engine company based in Santa Monica founded in 1998. Oingo focused on semantic searches rather than brute force string searches. - Oingo changed its name to Applied Semantics, which was then bought by Google for $102 million in April 2003, to replace a similar system being developed in house.
On Adwords, If you pass them wild stuff like regular expressions they sort of choke. I mean, they really really try to crank out the correct category. Then if they can't they fail with default categories. Sort of a "fail to zero" kind of thing . My default categories were "pimp your profile" and something about ringtones. If they are off base, and don't reflect the content of your site - they are easy to fix if they are totally off base. All you have to do is on the little "ads by google" link. Just click that, then at the bottom of the page. it says "Send google your thoughts on the ads you just saw".
I like adSense. I think its a fun thing to have on the site, whether you clickthrough or not. You get paid by a complex pricing model (based on a Vickrey second price auction, in that it commands an advertiser to submit a sealed bid not observable by competitors). For any given click received, advertisers only pay one bid increment above the second-highest bid - all of that being behind the scenes. One thing the adsense looks for is the phrase "click on my ads" . You are not allowed to say that. Check out my sponsored links. >:) Thats supposed to be ok.
. We can search for single terms, but we really can't search meaning yet. The basic context is beyond search right now. That is, the way that you and I are more or less communicating. (*looks in your eyes) . They did adSense in mySql. Tried to migrate it to Oracle but it got slower so they moved it back again because it didn't port right. There has to be something there. I mean. what exactly , would be such a big difference .. that has to tell you something about their SQL. I am not sure what.
So anyway. Just.. I mean. Sometimes when I do this I like the context it brings. You know? A little planet. A little star. What would the adwords say if they could hop into that context. O b a fine girl..kiss me.