I hope you understand: it is a core point that systems that answer questions must have the ability to internally represent the objects they assess in a way that allows reliability. Whatever the object.