Late January, we announced the kick-off of our 2010 Key Scientific Challenges Program, what we hope is a thought-provoking series of guest blog posts here on Yodel Anecdotal that offer a quick overview of these scientific challenge areas. Check out our last post on microeconomics and social systems. Today, we cover advertising, which is yet another important topic related to the Web. We’ve recruited Vanja Josifovski from Yahoo Labs to give you a glimpse of how significant advertising is to the online world. If you’ve seen Mad Men, you’ll know what it’s like to develop an infatuation with the 1960s advertising world – the fashion, the glamour, the old technology. While we’ve come a long way since then, advertising continues to remain a large part of our daily lives and a permanent fixture of our online experience. Here at Yahoo, we have a special appreciation for the undercurrents of online advertising because it supports a large swath of the Internet ecosystem. Generally, web advertising closely follows the same two major models of traditional advertising. First, there is direct advertising, which simply means that the goal of the ad is to get a direct response from the viewer. A good example is sponsored search where ads are shown beside a search result. The most difficult thing with this type of advertising is making sure the ads are relevant. If you searched for the word puzzle – did you mean jigsaw, crossword or Legos? These are complex issues we have spent a lot of time thinking about – and not in the Mad-Men-sit-back-with-whiskey way, but more of a microeconomics, super-computing way. And then there is brand advertising, which aims to create a favorable impression about a brand or a product. In this construct, ads are usually served up by graphical means, in what we call display ads on web pages. The challenge is matching the placement/context of the ad with behavior of the viewer, both of which also introduce a stunning number of variables and data points. In both modes – search and display – some complex learning algorithms are needed to provide relevant ads to a user. In a nutshell, it is this problem of relevance that makes advertising one of our Key Scientific Challenges. And it’s also central to the somewhat newer concept of computational advertising -- a scientific discipline we developed at Yahoo Labs that aims to formalize the problem of finding the best ad for a given user in a given context. It sounds logical enough, right? Can we really build a system that serves the most relevant ad to a user every single time? Can we bring advertising into the realm of the other "computational" sciences? Well, we think so and we’re working hard to tackle that kind of personal relevance. It’s simply going to take some hard core science, some persistence, and continued sharp thinking and hopefully more programs like those at universities (Stanford has a leg up) and our own Key Scientific Challenges to crack it. If you want to join in, here’s some fun reading to get you started:

• Translating Relevance Scores to Probabilities for Contextual Advertising Deepak Agarwal; Evgeniy Gabrilovich; Rob Hall; Vanja Josifovski; Rajiv Khanna, The 18th ACM Conference on Information and Knowledge Management (CIKM), 2009
• Robust Cross-Language Query Classification with External Web Evidence Xuerui Wang;Andrei Broder;Evgeniy Gabrilovich; Vanja Josifovski; Bo Pang, Proceedings of the 2nd ACM International Conference on Web Search and Data Mining (WSDM), 2009
• Context Transfer in Search Advertising Hila Becker;Andrei Broder; Evgeniy Gabrilovich; Vanja Josifovski; Bo Pang, The 32nd Annual ACM SIGIR Conference, 2009
  
  

Vanja Josifovski Yahoo Labs P.S. If you’re curious about just how big a deal Computational Advertising is, one of Yahoo’s foremost experts in the field, Andrei Broder, was just elected to the National Academy Engineering. That’s pretty good testimony that advertising is a little bit art AND science.

More information at:

Yodel Anecdotal