Enhancing the Power of a Decidable First-Order Reasoner

A major challenge in knowledge representation has been to devise reasoning mechanisms that are computationally feasible. The problem is that knowledge is usually incomplete and hence calls for very expressive representation languages like that of first-order logic, yet reasoning about incomplete knowledge is undecidable when based on classical logic. Over the past decade there have been several semantic approaches defining decidable forms of first-order reasoning. The computational gain, however, came at the price of losing too many useful inferences. In this work we take one of these existing weak reasoners and extend its power without losing decidability by moving from an unsorted to a sorted logic. In contrast to similar work by Frisch, we are not limited to formulas in a certain normal form and our approach extends to introspective reasoners as well.