Tabling in contextual abduction with answer subsumption

Abduction is a form of logical inference that seeks out best explanations for a given observation. Abduction has already been well studied in the field of computational logic, and logic programming in particular. In contextual abduction, explanations obtained within one context may also be relevant in different contexts. In such contextual abduction, explanations thus can be reused with little cost. When abduction is realized in logic programming, one can reuse previously obtained explanations from one context to another by benefiting from a logic programming's feature called tabling. In this paper, we revisit tabling in contextual abduction and improve this technique with an advanced tabling feature of XSB Prolog, viz., answer subsumption. The employment of answer subsumption in this technique is important, when one is interested in obtaining minimal explanations for an observation. It also helps avoid tabling too many and large explanations for a given observation, which may fail contextual abduction in practice, as it requires too many resources before being able to return a solution. We provide a protoytpe, tabdual
⁺
, of this improved technique as a proof of concept. Our experiments, both in artificial and real world cases, show that Tabdual
⁺
correctly returns minimal explanations, while the cost of their computation is greatly reduced.