Ontology-based data access (OBDA) is a useful approach to gain access to existing structured data in an easy and intuitive manner, abstracting from the representation of the data and thus releasing from the burden of having to deal with technical details of the data structuring. By using the approach of ontology bootstrapping, an ontology reflecting the structure of the data can automatically be created by examing the data source. This allows to benefit from ontology-based data access concepts without having to forbear from the advantages regarding performance and data consistency which is offered by relational database systems. Furthermore, establishing a system providing a so-called “virtual RDF view”, data from relational sources can be accessed and even updated by querying or modifying the ontology while remaining in the database. However, the creation of an ontology from another structured data source is a complex task that involves several tools, formats and languages. Also taking changing input data into account, the process of ontology bootstrapping becomes a significant engineering task in its own right. The approach of introducing OBDA specifications proposed by Skjæveland et al. address this issue by concentrating all relevant information on the bootstapping process in one place, allowing for systems which drive the entire tool chain based on this information and thus rendering the use of ad-hoc scripts unnecessary. A concrete format for serializing such an OBDA specification for storage or exchange, however, is not provided. The introduction of such a format is part of this thesis. After the task it accomplishes, the format was named “OBDA Specification Language” (OSL). Furthermore, this thesis describes the design and implementation of db2osl, a software which is able to, in turn, automatically bootstrap an OBDA specification from a relational database schema. Finally, it gives an explanatory description of both the process of ontology bootstrapping and the process of an OBDA specification and proposes a refined scheme for generating IRIs for the entities of the resulting ontology.