When scientists or engineers are looking for information in document collections, or on the web, they generally have a precise objective in mind. Instead of looking for documents "about a topic T", they rather try to answer specific needs such as finding the definition of a concept, finding results for a particular problem, checking whether an idea has already been tested, or comparing the scientific conclusions of two articles. This paper presents an indexing model which includes the decomposition of documents into fragments that will correspond to discourse elements (definition, hypothesis, method, result, etc.). The division of documents into fragments should allow scientists to retrieve more relevant information and to make queries more precise. Each type of discourse element will be modeled by defining specific characteristics. The representation model will allow operations that merge or combine retrieved documents or fragments. We will propose an interface that allows strategic reading such as parallel reading of fragments.