Relevant Backtracking: Improved Intelligent Backtracking Using Relevance

This work investigates and develops a backtracking algorithm with a novel approach to enumerating and traversing search space. A basic logical framework inspired by relevant logics is presented, highlighting relationships between search and refutation proof construction. Mechanisation of a relevance aware Davis Putnam Logemann Loveland procedure is investigated, and this yields an intelligent backtracking algorithm with abilities similar to other mechanisms including extended freedom in manipulating search space or rearranging refutation proof construction. Simplicity is achieved by a separation of concerns of the underlying logic and the construction of a sound and complete algorithm. The key advantage in the method is that it captures the notion of proof construction and relevant causality, and empirical analysis shows that this is an effective approach.