Semantic Web technologies facilitate data integration over a large number of sources with decentralised and loose coordination, ideally leading to interlinked datasets which describe objects, their attributes and links to other objects. Such information spaces are amenable to queries that go beyond traditional keyword search over documents. To this end, we present a formal query model comprising six atomic operations over object-structured datasets: keyword search, object navigation, facet selection, path traversal, projection, and sorting. Using these atomic operations, users can incrementally assemble complex queries that yield a set of objects or trees of objects as result. Results can then be either directly displayed or exported to application programs or online services. We report on user experiments carried out during the design phase of the system, and present performance results for a range of queries over 18.5m statements aggregated from 70k sources.