Ed: Another blind men with the elephant problem. Contextual or behavioral inference depends on artifical intelligence rules. Cute stories aside, these methods don't scale.
from Lightspeed Venture Partners Blog by
IEEE Spectrum notes three new patent applications from Google that suggest how Google hopes to improve targeting for social media advertising:
The patents—Open Profile Content Identification, Custodian Based Content Identification, and Related Entity Content Identification—and the algorithms behind them would let Google find patterns in users’ profiles, pages, and friend lists in order to better target ads to them. Ideally, they would make the users more likely to click through.
Google’s Related Entity patent, for one, involves prying information from a user’s list of friends or user groups. But Google is not alone in this field. In June the social ad firm SocialMedia Networks said it had invented an algorithm called FriendRank that also scours a user’s friendship lists for friends whose names might be dropped in a targeted ad.
The Open Profile and Custodian patents would mine data from, say, a MySpace user’s profile and the profile of the MySpace page the user is visiting. The Open Profile patent, for instance, would consider a user profile like “I really enjoy hiking, especially long hikes when you can camp out for a few days. Indoor activities don’t interest me at all, and I really don’t like boring outdoor activities like gardening.”
Using smart language-processing algorithms to detect the user’s sentiments (“enjoy” or “don’t like” near “hiking” or “gardening”) and other linguistic cues, the system would then potentially serve up active outdoor sports-related ads to this user but avoid ads about more hobbyist-oriented activities.
Google is continuing to apply the adsense paradigm of contextual targeting, but expanding the definition of “context” to include the friend networks and the declarations of interests that are common to social network profile pages. MySpace’s hypertargeting is a similar approach.
Facebook’s engagement ads are a markedly different approach, charging for actions rather than using targeting to lift CPMs within display advertising. I think engagement ads could be a very interesting approach that takes advantage of the native behavior of “user affiliation” withing social networks.
I’m indifferent to which standard wins, but I do want a standard to emerge. We need a standard in social media advertising to unlock further scalable growth in this industry.
A flurry of reports yesterday suggest that Twitter 
TheRarestWords then tries to tap into crowd intelligence by letting anyone add a 100-character definition for each keyword, which could give it a semantic edge in trying to categorize each site. This could also be gamed pretty easily, but this looks to be just a Web project at this point. It could also be used to create a Wiki dictionary like
Is Google done with its famous motto, "Don't be evil?" TechCrunch's Michel Arrington 
"
Facebook
LinkedIn
about adEcon
MarSP
rich media
top
blog
newspaper
social network