The Future of Data Integration

Renée J. Miller

Abstract: The value of data explodes when it is integrated. In this talk, I present some important innovations in data integration over the last two decades. These include data exchange [1], which provides a foundation for reasoning about the correctness of transformed data, and the use of declarative mappings in integration [2]. I discuss how data mining has been used to facilitate data integration, including constraint discovery [3], mapping discovery [4], and in schema discovery to combat database decay and facilitate integration [5,6]. I present some important new data integration challenges that arise in data science. These include the use of mining for query and visualization recommendation over massive data lakes [7] and data set search, finding datasets of interest at interactive speeds [8].