Justification Based on Program Transformation

Justification is to give evidence, in terms of a proof, for the truth value of the result generated by query evaluation of a logic program. In an earlier work we presented algorithms for justifying logic programs. It was based on using tabling for evaluating the program and post-processing the memo tables created during evaluation. Justification built in this fashion does not compromise the performance of query evaluation in the sense that it is completely decoupled from the query evaluation process and is done only after evaluation is completed. However, this algorithm is based on metainterpreting the memo tables and clauses of the program. This is a major source of inefficiency. In this paper we present a new justification scheme based on program transformation. Justification of a true literal is generated during query evaluation of the transformed program. To justify a false literal, we generate the dual definition for each predicate defined in the program. A new program is then obtained based on those dual predicates and the original program. Query evaluation on this new program yields justification for the false literal. We provide experimental evidence to show that justification using program transformation outperforms the previous techniques in both efficiency and scalability.