We present a modular breakdown of data integration tasks and the results of a survey on the distribution of effort among those tasks. The modularization aids in project planning and enables portions of the work to be allocated flexibly among various human specialists, and also to automated tools. The survey results are useful for determining: (1) what are the highest value research problems to tackle? (2) where should large enterprises focus their investments in data integration tools?, and (3) how should a data integration project manager allocate people and schedule?
Leonard J. Seligman, Arnon Rosenthal, Paul E. Lehn