The Problem:
A data-rich world has a serious downside: once you find some of what you want, how do you find more? PageRank, collaborative filtering, and other approaches strive to solve this problem, but from Googlewhacking to Amazon customers whose other interests you don't care about, each one is flawed or even outright exploitable.

Humans may not be able to say “why” they like something, but they can identify things they do and don't like. How can we use this to help people fish new things they'll like from a sea of data?

The Closed Solution:
Existing web services, like OkCupid and Pandora, collect information about resources and show you new resources that are similar to resources you've already identified. OkCupid asks you questions about yourself, and then shows you people who have answered those questions in a similar way. Pandora asks you for a band or song, and then plays you new music by artists they've identified as sharing qualities with the music you already like. That's great, but does every service have to develop its own solution?

It shouldn't—and now, it doesn't.

The Open Source Solution:
This talk presents the theory behind and the practical use of Query by Example, a free-as-in-freedom add-on to PostgreSQL which lets users specify a few examples of data they like and data they don't like, then retrieves unseen data points which match their preferences. It will focus on rapid development of new applications.