Although statistical graphs are ubiquitous, few techniques and standards exist to exchange, search and query these graphical representations. The gSem project aims to improve human-graph interaction by developing new approaches to manage statistical graph knowledge. Two specific objectives of this project are addressed in this paper: (1) improve the efficiency of searching statistical graph knowledge and (2) facilitate sophisticated question answering about statistical graph knowledge. In particular, semantic annotation and query answering across statistical graphs with OWL ontologies are described.