Any Machine Learning engineers here willing to take the challenge of generating a prolog engine from the natural language descriptions of constraints?
Those kinds of issues don't prevent someone from whipping up something, but guarantee it'll have terrible accuracy. Anybody trying to use it would probably spend more time debugging the generated Prolog than they would just writing the Prolog themselves.
PL research isn't really focused on translation anyway; the current trends I've seen are block-structured editors where you would snap things together to form a constraint, probably using a formal mathematics-inspired syntax similar to the LaTeX stuff that this catalog is using, and using SMT solvers to avoid having to deal with the complexities of procedural languages.