This thesis presents an automated, data-driven integration process for relational databases. Whereas previous integration methods assumed a large amount of user involvement as well as the availability of database meta-data, we make no use of meta-data and little end user input. This is done using a novel join and translation finding algorithm that searches for the proper key / foreign key relationships while inferring the instance transformations from one database to another. Because we rely only on the relations that bind the attributes together, we make no use of the database schema information. A novel searching method allows us to search the database for relevant objects without requiring server side indexes or cooperative databases.