In this paper some new theoretical results from bilinear sytems theory are presented. New formulations of the concepts of span-reachability and observability are given. Based on these investigations standard results from linear systems theory, such as canonical forms and a rank test for the state space decomposition are transferred to bilinear systems theory. In particular, a procedure is outlined which, using recent theoretical results, enables the determination of system parameters from measured input/output data. This approach reduces the realization and identification problem of bilinear systems to that of matching a finite sequence of input/output parameters. On this basis it is possible to develop, for bilinear systems, algorithms analogous to the ones originated from well-known results for linear systems. To illustrate the achieved algorithms, the results of the identification of bilinear models for a translational hydraulic drive out of measured input/output data are presented.