When managing a virtual observatory we have to cope with the problem of providing an uniform access to data and services. But we must also take into account the particular groups of users for which these data and services are provided. Thus, it becomes crucial to have a visibility on the manner in which each group of users is using these data. In our approach we consider that the main output of an astronomical data use lies in papers published by researchers and that the trace of this use is valuable to tackle the way data are used.
On a theoretical point of view we use data mining tools in order to identify the different communities of users of a given set of data. We then use visualisation tools to identify the way each community uses these data over time.
On a practical point of view we start with different set of data managed by the CDPP (Plasma Physic Data Center). We then locate the articles which correspond to each set of data in the ADS bibliographic database. Then, we display a synthetic and diachronic view of the history of the publication involving each set of data. Finally, we give a framework in order to interpret this view through different parameters: duration of use, density of use, use burst, etc.