RDF to ABNF

Has anyone had any experience creating sets of Augmented Backus-Naur Format resources based on RDF data primed from a rdf dataset, with a view to using this the basis of a Natural Language Interface.

Example:

gr:ProductOrServiceModel a rdfs:Class.

grammar:Product-pattern a grammar:NLPRule;
       grammar:abnf-pattern "product-pattern = product-adjective product"
       grammar:sparql-criteria-builder grammar:Cheapest-product-criteria-builder.

grammar:Cheapest-product-criteria-builder a grammar:sparql-criteria-builder;
      sparql:triple-pattern "?binding a gr:ProductOrServiceModel";
      sparql:sequence-modifier "ORDER BY ?price";
      sparql:sequence-modifier "LIMIT 10".

grammar:Product-adjective a grammar:NLPRule;
       grammar:abnf-pattern "product-adjective = 'cheapest'".

Loop over the products in the datastore:

grammar:Product-noun a grammar:NLPRule;
       grammar:abnf-pattern "product = 'Nokia Lumia 800'".

I am aware that there is probably an rdf vocabulary for SPARQL and that the above is an example, not really a solution, but I would be interested if anyone has any real solutions or approaches that are:

1. Extensible across different domains
2. Manageable, measurable and efficient (i.e least manual priming required)

Additionally, I won't be offended if nobody understands the question (I don't think I have explained the problem very well).

UPDATE: There may also be some aspects of SPIN that might make this more manageable