<algorithm> An algorithm for proving a goal by recursively breaking it down into sub-goals and trying to prove these until facts are reached. Facts are goals with no sub-goals which are therefore always true. Backward training is the program execution mechanism used by most logic programming language like Prolog.
Opposite: forward chaining.
Try this search on Wikipedia, OneLook, Google
Nearby terms: Backus-Naur Form « Backus Normal Form « backward analysis « backward chaining » backward combatability » backward compatibility » backward compatible