Keyword Search: Integration on Demand

A recent focus in data integration has been on how to enable “average” (non-database-expert) users to pose ad hoc queries over structured, integrated data. The approach is to use a familiar interface: keyword search. Here, however, keyword search does not merely match against single documents or objects. Instead, the set of keyword terms describes a set of concepts in which the user is interested. The data integration system is tasked with the job of finding a way of relating the tables or tuples relating to these concepts, e.g., through a series of joins. In this chapter, we first describe the abstract problem of keyword search over structured data, then we present several popular algorithms for returning ranked results efficiently. Finally, we describe some of the issues in implementing keyword search over data integration applications.