How to setup inference
Within RDFScape, ontologies that are queried or visualized are not only the ontologies "loaded" into the system. They are processed through a configurable inference step that allows to specify standard and custom inference rules. In this way, ontologies can be "interpreted" and transformed to follow some analysis needs. Some examples can be found in the Documentation page.
1 Selection of the "Knowledge Engine". Jena is the only option provided at the moment.
2 Select the level of standard entailments that you wish to be computed. Options are:
- None No reasoninig performed (other than selected inference rules)
- RDFS-1 Uses the Jena transitive reasoner ( more details here). Custom inference rules are applied "on top" of these entailments.
- RDFS-2 Uses the Jena rule resoner with partial entailments for RDFS (more details here). Custom inference rules are applied "at the same time as" these inference rules.
- OWL-1 Uses the Jena rule reasoner with partial entailments for OWL, corresponding of the OWL-Micro specs (more details here). Custom inference rules are applied "at the same time as" these.
- OWL-2 Uses the Jena rule reasoner with partial entailments for OWL, corresponding of the OWL-Mini specs (more details here). Custom inference rules are applied "at the same time as" these.
- Pellet-fast Uses Pellet. The fast option is an attempt to compute less entailments, to gain the same perceived speed as in the command line pellet utility. In this case only the subsumption computation from Pellet is used, and its result (classes and properties) is added to the original model and passed to an RDFS model (toghether with custom inference rules) for transitive types resolution.
- Pellet Uses Pellet. This is significantly slower than the command-line equivalent. In fact, routing Pellet thorugh Jena issues a full expansion of entailments. Therefore this mode should be used (at the moment) only for small ontologies. Custom inference rules are applied "on top" of the entailments genereated by Pellet.
- DIG Asks for a URL where a DIG reasoner can be found. This feature is under testing. Was verified with Pellet, failed with Fact++.
3 Selecting an inference rule "title" displays the relative rule-text in the panel on the right. Only rules whoe checkbox are selected are included in the reasoninig step (all are saveed in analysis contexts).
4 Specify the text of the current rule. Syntax and examples ofr rule editing can be found here. The title of the rule is automatically parsed and entered as a label in the rule lists. Rules are memorized (and parsed) only after commit is selected. Before that, just visualizing another rule will delete all uncommitted changes at the current one. At this time, prefixes in rules are not supported. note that all the rules should be backward!
5 Add a new rule entry. This doesn't imply commit of the current rule (commit before!)
6 Commit the current rule. The rule is parsed and errors are reported. Only when committed the rule remains in memory. Saving the current analysis context at this point will save it to disk.
7 Deletes the selected rule.
8 Reason. Issues reasoning with the current features selected. This is automatically called the first time ontologies are loaded, but must be invoked by the user on order for change in configuration to take effects. Effects of the new settings will be directly avalilable in RDFScape.
9 Post Rule paramaters. Currently unsed.
10 Jena Rule syntax. Currently only Jena is supported.