We give an overview of the SIRUP (Semantic Integration Reflecting User-specific semantic Perspectives) approach to semantic data integration that takes into account heterogeneity of data receivers. Our goal is to provide means that allow data from heterogeneous sources to be integrated in a way that it perfectly fits to a particular users information needs, emphasizing his individual way to perceive a domain of interest. To achieve this, we propose to use a semantic multidatasource language to declaratively manipulate so-called IConcepts; these are basic conceptual building blocks to which attribute data that refers to the same real-world concept is linked by data providers. We provide explicit, queryable semantics by connecting IConcepts to concepts of ontologies. Additionally, pre-integrating data on a conceptual level through IConcepts shields SIRUP end-users from low-level heterogeneity and technical details of underlying data sources.