Aardvark Top Query Categories

Aardvark Example Query

Social Search

2007 - present

Much research in Information Retrieval (and indeed, most development of IR systems in practice) assume an information need and a collection of documents, and further assume that trust in a document should be based on authority. This clearly has worked well for a broad range of information needs.

But in the offline world, there is a large class of knowledge that is disseminated socially, and the retrieval task consists of finding the right person, rather than the right document, to satisfy the information need. Further, in many of these cases, trust is based on intimacy rather than authority.

In this model, the combination of real-time responses by socially proximal human responders allows for questions that are natural language, rich in detail, and subjective in nature. (For example: "I'm going on a second date with a cool, spontaneous, funky girl on Saturday night in San Francisco. Do you recommend any restaurants for dinner?")

Our research involves building IR systems to answer questions like this by mimicking what people do offline while decreasing the associated social burden. We focus on algorithms, interfaces, systems, for social search.

This is joint work with Damon Horowitz, Max Ventilla, and the entire Aardvark team.


  • The Anatomy of a Large-Scale Social Search Engine. Nineteenth International World Wide Web Conference, April 2010. (with Damon Horowitz).

  • A Bayesian Derivation of Personalized and Social Search Stanford University Technical Report, October 2011. (with Damon Horowitz).


  • Aardvark. Technical Advisor. Social search engine acquired by Google in February, 2010.