When scientists are looking for information in document collections, they generally have a precise objective in mind: They want to find all the definitions of a concept, all the results for a particular problem, or to compare the scientific conclusions of multiples articles. In this research, we have developed a user-centric annotation model based on discourse elements. The model has been formally defined as an OWL ontology and used to annotate a corpus of scientific articles in gender studies. We show how complex queries, proposed by scientists, can be expressed in this model and solved by a description logic reasoner.