Incremental and Persistent Reasoning in FaCT++

Abstract. Reasoning in complex DLs is well known to be expensive. However, in numerous application scenarios, the ontology in use is ei-ther never modified at all (e.g., in query answering), or the amount of updates is negligible in comparison with the whole ontology (e.g., minor manual edits, addition of a few individuals). In order to efficiently sup-port these scenarios, FaCT++ implements the following two techniques: persistent and incremental reasoning. In persistent reasoning mode, after classification, the reasoner saves its internal state, including computed information (e.g., concept taxonomy) on a persistent medium; the next use of the ontology will not require classification to be performed from scratch, but just reading an input file. In incremental reasoning mode, the reasoner is notified of a change and identifies a (usually small) portion of its internal state that is affected by the change. This is the only part that requires recomputation. This approach can lead to greater overall efficiency, when compared with having to reload and reclassify the whole ontology.