What Is the Best Thing to Do Next? A Tutorial on Intensional Data Management

We call data intensional when it is not directly available, but must be accessed through a costly interface. Intensional data naturally arises in a number of data management scenarios, such as crowdsourcing, Web crawling, or ontology-based data access. Such scenarios require us to model an uncertain view of the world, for which, given a query, we must answer the question “What is the best thing to do next?” Once data has been retrieved, the knowledge of the world is revised. This tutorial is an introduction to intensional data management, with a review of the solutions brought in various areas of data management and machine learning, and of some challenging open problems.