The web of data is a large collection of various data from different domains, these data are distributed on several sources which are heterogeneous and managed independently. Hence, a user may need to query multiple data sources to aggregate pieces of information. In this context, our work aims to set up a system that integrate fragments of information from distributed and independent data sets by providing a single interface to the user. In this paper, we propose a system designed for querying the web of data without having a prior knowledge of the sources contributing to the response. This system adopts a solution for selecting relevant sources in order to decrease network traffic so that the system becomes less dependent on the quality of the connection flow.