This paper presents exploratory work on the use of semantics in Document Image Analysis. It is different than existing semantics-aware approaches in the sense that it approaches the problem from a a very domain specific angle, and tries to incorporate an open model based on a reduced ontology. As presented here, it consists of enhancing an existing platform for Document Image Analysis benchmarking using off-the-shelf tools. The platform on which it is based hosts a wide variety of image interpretation algorithms as well as a wide range of benchmarking data. These data are stored in a relational database, as well as their type definition, the association between data and algorithms, etc. This work tries to provide an experimental indication whether ontologies and automated reasoning can provide new or alternative ways to extract relations among different stored facts, or infer dependencies between various user-defined types, based on their interactions with algorithms and other types of data.