As more information becomes available, the ability to
quickly incorporate new and diverse data sources into existing
database systems becomes critical.  Schema coercion addresses this
need by defining the mapping between databases as a collection of
mappings between corresponding constructs.  This work defines a
comprehensive schema coercion tool: it transforms schemata into
corresponding ER representations, identifies correspondences between
them, and uses these correspondences to generate a program that
automatically transfers data between the databases. In addition to
creating a useful tool, this work addresses the significant
theoretical problems associated with resolving representational and
semantic conflicts between heterogeneous data sources.  The approach
advocated by this dissertation associates confidences with
correspondences, and meta-information with schemata.  This approach
has successfully reduced the amount of interaction required to define
several coercions, including a complex coercion between diverse
genetics databases.