Web Reasoning Using Fact Tagging
Mehdi Terdjimi, Lionel Médini, Michaël Mrissa · 2018
Today's Web applications tend to reason about cyclic data (i.e. facts that re-occur periodically) on the client side. Although they can benefit from efficient incremental maintenance algorithms capable of handling frequent data updates, existing rule-based algorithms cause successive re-derivations of previously inferred information. In this paper, we propose an incremental maintenance approach for rule-based reasoning that prevents successive re-computations of fact derivations. We tag (i.e. annotate) facts to keep trace of their provenance and validity. We compare our solution with the DRed-based incremental reasoning algorithm and show that it significantly outperforms this algorithm for fact updates in re-occurring situations, to the cost of tagging facts at their first insertion. Our experiments show that this cost can be recovered within a small number of cycles of deletions and reinsertions of explicit facts. We discuss the utility and limitations of our approach on Web clients and provide implementation packages of this reasoner that can be directly integrated in Web applications, on both server and client sides.