In the Virtual Extension

Viewpoints article — DOI: 10.1145/2018396.2018423
Information Seeking: Convergence of Search, Recommendations, and Advertising

Hector Garcia-Molina, Georgia Koutrika, and Aditya Parameswaran

All of us are faced with a "deluge of data" in our workplaces and our homes: an ever-growing World Wide Web, digital books and magazines, photographs, blogs, tweets, email messages, databases, activity logs, sensor streams, online videos, movies and music, and so on. Thus, one of the fundamental problems in computer science has become even more critical today: how to identify objects satisfying a user’s information need. The goal is to present to the user only information that is of interest and relevance, at the right place and time.

At least three types of information-providing mechanisms have been developed over the years to satisfy user information needs: A search mechanism takes as input a query that describes the current user interests; a recommendation mechanism typically does not use an explicit query but rather analyzes the user context and if available, a user profile; an advertisement mechanism is similar to a recommendation mechanism, except that the objects presented to the user are commercial advertisements, and financial considerations play a central role in ranking.

The authors argue that these mechanisms are not that different to begin with, and designing these three mechanisms making use of this commonality could lead to significant benefits. All three mechanisms share the same common goal: matching a context (which may or may not include an explicit query) to a collection of information objects (ads, product descriptions, Web pages, and so forth). The way context can be described varies, but there is no fundamental reason why one object type needs to be associated with one mechanism type. Similarly, the way matches are determined and ranked varies, but again, there is no fundamental reason why approaches cannot be used across mechanisms.

After reviewing some of the key concepts that historically evolved for search, recommendations, and advertisement, the authors explain why they believe a convergence of these technologies would be more useful today than in the past.