Adpinion: a recommender system for online ads
I learned about Adpinoin yesterday while surfing around. They are a recommender system for advertisements. Essentially what happens is that users rate ads that they are show via a ratings widget on the ad itself. As you vote on ads the system builds up a profile for you, ads that you like remain visible, ads that you don’t like are remove and the system tells you “sorry”. As you view and rate ads on the network it gets smarter about what ads to place in front of you based on what you and other like you vote on. From their overview page it appears they are using a clustering model for their collaborative filtering technique. This seems like a very nice way to model the user rating data.
This is really a neat idea and one that I have thinking about these past couple of weeks. I wonder if a better approach for gathering ratings on the ads would be using actual clicks on the ads as votes. If the user clicked on the ad then it is a thumbs up otherwise it is a thumbs down. Or perhaps use both, and have a graduated system where thumbs down = 0, no click = 1, click = 2, thumbs up = 3. This might offer more data to the system and help build better clusters. Also this would allow them to start gathering data on text ads instead of just banner ads that most people already ignore, or have their web browser filter out.
In any event I think this is a great first step and it will be interesting to see if these ads pop up on some of the websites that I frequent.
Share This-->