The convenient availability of information is an essential factor in science and business. While internet technology has made large amounts of data available to the general public, the data is largely provided in human-readable format only. New technologies are now making direct access to millions of structured or semi-structured databases possible, but only through integration of these data sources maximum benefit can be gained. Traditional approaches to information integration, which involve human development teams and work in a controlled environment with a stable set of data sources, are not applicable due to the dynamic nature of such an environment. Therefore a higher degree of automation of this process is required. We present the PALADIN project (Pattern-based Architecture for LArge-scale Dynamic INformation integration), that uses machine-understandable patterns to capture and apply expert experience in the integration planning process.