Scientific Challenges in Contextual Advertising

Abstract

Online advertising has been fueling the rapid growth of the Web that offers a plethora of free web services, ranging from search, email, news, sports, finance, and video, to various social network services. Such free services have accelerated the shift in people’s media time spend from offline to online. As a result, advertisers are spending more and more advertising budget online. This phenomenon is a powerful ecosystem play of users, publishers, advertisers, and ad networks. The rapid growth of online advertising has created enormous opportunities as well as technical challenges that demand computational intelligence. Computational Advertising has emerged as a new interdisciplinary field that studies the dynamics of the advertising ecosystem to solve challenging problems that rise in online advertising.

In this talk, I will provide a brief introduction to various forms of online advertising, including search advertising, contextual advertising, guaranteed and non-guaranteed display advertising. Then I will focus on the problem of contextual advertising, which is to find the best matching ads from a large ad inventory to a user in a given context (e.g., page view) to optimize the utilities of the participants in the ecosystem under certain business constraints (blocking, targeting, etc). I will present a problem formulation and describe scientific challenges in several key technical areas involved in solving this problem, including user understanding, semantic analysis of page content, user response prediction, online learning, ranking, and yield optimization.