Abstract : Informational objects are support or mediator used to improve perception of information when people are working with computer. They must be simple texts, combined or not with fixed or animated picture. They must be produced manually by one or many authors, or dynamically by information retrieval systems like answers are. These objects must be viewed as aggregates of indexed elements, generally called fragments, stocked in databases. Strategies of constructions are varying from systems to systems so objects are very different from each others but we can describe formerly them by three characteristics; there degree of aggregation, scheduling and repetition of fragments. Developers and ergonomicists very often use these three characteristics to improve use, relevance and efficacy of systems. Paradoxically, because any formalism as been defined, they are not taken into account to evaluate or compare results constructed. In this paper we propose to present a mathematical formalism based on the concept of semi-order and possible to use in the case of IR results. Then we present how it can be write with the fuzzy set theory and used to construct similarity measure taken into account the presence and decrease rank of documents. To finish we present other contexts of use and research perspectives.