The aim of this paper is to illustrate how a probabilistic Description Logics, called DL _P, can be exploited for reasoning about information sources characterized by heterogeneous formats and structures. The paper first introduces DL _P syntax and semantics. Then, a DL _P-based approach is illustrated for inferring complex knowledge assertions from information sources characterized by heterogeneities in formats and representational structures. The thus obtained complex knowledge assertions can be exploited for constructing a user profile and for improving the quality of present Web search tools.