Statistical language modeling combining N-gram and context-free grammars

Linguistic structure in the form of a partial-coverage phrase structure grammar is combined with statistical N-gram techniques. The result is a robust statistical grammar which explicitly incorporates linguistic and semantic structure. This approach makes it possible to model carefully those parts of the input that are important for an application and to use robust techniques that provide a full-coverage statistical language model. This approach is being applied to the recognition of air-traffic-control transmissions, and it has already been shown that a simpler hybrid approach is useful.<>