[Back to DRAOn]
Technical Details of DRAOn
DRAOn uses the OWL API to manipulate OWL ontologies and the Alignment API to deal with aligments. In addition, DRAOn is based heavily on HermiT reasoner that is charged in reasoning tasks over simple ontologies.
In the version 1.5, we implement optimizations for reducing the number of configurations to be considered and for improving communication protocol between local reasoners.
In addition, the entailment service for the IDDL semantics is introduced for some concept axioms.
- DRAOn allows for reasonning with a network of aligned ontologies. In this context, DRAOn considers alignments as knowledge fragments independent from ontologies. If alignments would result from ontology reconciliation for establishing interoperability, reasoning with a network of aligned ontologies allows us to check whether this reconciliation process guarantees consistency of the whole network under a certain semantics.
- DRAOn supports the IDDL semantics for a network of aligned ontologies. The IDDL semantics is weaker that the DL semantics, that means, there is an IDDL consistent network of aligned ontologies that is not DL consistent. This results from the fact that IDDL considers alignments as "meta-knowledge" with respect to ontologies. This would be more plausible in several cases where disjointness relationships between classes or properties should not be propagated from ontologies to alignments. Thanks to IDDL, we can define and check global consistency of a network of aligned ontologies without requiring that all formalisms used in ontologies are the same.
- DRAOn supports a distributed reasoning under the IDDL semantics that makes possible to distribute reasoning tasks over local ontologies to different locations (local reasoners). Although the number of messages exchanged between a global reasoner and local reasoners may be very high, an implementation taking advantage of parallelization and offline computation can improve dramatically response time to user's requests.
In this version, we have developed and implemented new techniques whose behaviour is more goal-oriented than those implemented in early versions. The main ideas are:
- Configurarions (corresponding to sets of axioms) are built in an incremental way, and reused if possible results of previous checks.
- For each configuration, we check entailment rather than consistency. This allows for putting forward eventual backtracking points.
- Communication protocole between global and local reasoners is parallelized, that means, requests are sent to local reasoners in a broadcasting way rather than the sequential one.
The inference services supporting the IDDL semantics are parallelized and implemented in a distributed way in this version. When defining a network of aligned ontologies, a global reasoner that and several remote local reasoners are created. The global reasoner uses several threads to manage communication between the remote reasoners and itself. We have to use sockets to establish the communication between global and local reasoners instead of OWLink. This is due to effciency question and an intrinsic characteristic of IDDL.
Instead of allowing for all disjointion axioms in alignments, we implement entailment services by using the fact that the entailments $O_i \cup A \models C(a)$ and $O_i \cup A \models C \sqsubseteq D$ can be achieved by propagating disjointness axioms to alignments and non-emptiness axioms to local reasoners.
$IUT de Montreuil, Universite de Paris8 $