Reflections on Schema Mappings, Data Exchange, and Metadata Management

Abstract

A schema mapping is a high-level specification of the relationship between two database schemas. For the past fifteen years, schema mappings have played an essential role in the modeling and analysis of data exchange, data integration, and related data inter-operability tasks. The aim of this talk is to critically reflect on the body of work carried out to date, describe some of the persisting challenges, and suggest directions for future work. The first part of the talk will focus on schema-mapping languages, especially on the language of GLAV (global-and-local as view) mappings and its two main sublanguages, the language of GAV (global-as-view) mappings and the language of LAV (local-as-view) mappings. After highlighting the fundamental structural properties of these languages, we will discuss how structural properties can actually characterize schema-mapping languages. The second part of the talk will focus on metadata management by considering operators on schema mappings, such as the composition operator and the inverse operator. We will discuss why richer languages are needed to express these operators, and will illustrate some of their uses in schema-mapping evolution. The third and final part of the talk will focus on the derivation of schema mappings from semantic information. In particular, we will discuss a variety of approaches for deriving schema mappings from data examples, including casting the derivation of schema mappings as an optimization problem and as a learning problem.