Various issues in the implementation of generalized LR parsing with probability are discussed. A method for preventing the generation of infinite numbers of states is described and the space requirements of the parsing tables are assessed for a substantial natural-language grammar. Because of a high degree of ambiguity in the grammar, there are many multiple entries and the tables are rather large. A new method for grammar adaptation is introduced which may help to reduce this problem. A probabilistic version of the Tomita parse forest is also described.