Improving schema mapping by exploiting domain knowledge

This dissertation addresses the problem of semi-automatically creating schema mappings. The need for developing schema mappings is a pervasive problem in many integration scenarios. Although the problem is well-known and a large body of work exists in the area, the development of schema mappings is today largely performed manually in industrial integration scenarios. In this thesis an approach for the semi-automatic creation of schema mappings based on a central ontology is developed. The central ontology stores domain as well as integration knowledge. Using this ontology as a basis, schema mappings of high quality can be created automatically. In particular, this thesis addressed the following questions: • Which knowledge is necessary to create schema mappings? It is obvious that schema information alone is not sufficient to create mapping between complex schemas. The thesis analysis which types of background knowledge are required in order to create a correct mapping between two schemas. This analyses of required background knowledge is not particularly focused towards the automatic creation of schema mappings but also applies to the manual development of schema mappings. • How can the necessary background knowledge be collected? The background knowledge required to create schema mappings is usually not easily accessible. As a result, two approaches for the collection of background knowledge are developed and evaluated in this thesis. A particular focus is placed on enabling the non-intrusive collection of the background knowledge to enable the application of the approaches in productive environments. • How can background knowledge be exploited by an automated approach? In order to exploit background knowledge during the mapping creation, it needs to be available in a machine interpretable format. The thesis shows an approach for the modeling of domain and integration knowledge in an ontology and, based on this, develops a schema mapping approach capable of exploiting the modeled knowledge during the automatic schema mapping creation.