Abstract Meaning Representation (AMR) is a type of semantic graph, which is a convenient and popular way of representing natural language. Linguistic concepts are modeled as nodes with edges between them describing their relationships. A Graph Extension Grammar (GEG) is a type of graph grammar that can generate semantic graphs akin to AMR. The aim of this thesis is to explore the limitations and suitability of using graph extension grammarsfor semantic generation. This is done by modeling a single GEG after an AMR corpus.A large portion of this thesis is focused on generic structures in AMRs, and how tomodel them in a GEG. Further improvements to the formalism are also presented. The conclusion states that the GEG formalism is suitable for semantic graph generation and that it is possible to generate a corpus using a single GEG. However, large corpora may be difficult and time-consuming to model due to complex reentrancies.