Protégé DL-Learner Plugin
The Protégé 4 DL-Learner plugin allows to learn equivalence and super class axioms based on the instance data in the ontologies loaded in Protégé. It integrates seamlessly in Protégé.
You can either download the plugin or install directly from Protégé 4 itself: Go to “File“ => “Preferences“ => “Plugins“, click on “Check for Downloads“ and select the DL-Learner plugin.
The following screen cast shows how to activate the plugin in Protégé and use it:
The plugin was created by Lorenz Bühmann, Christian Kötteritzsch and Jens Lehmann. If you have any questions or comments, please contact me.
- learns definition and super class axioms based on existing instance data in an ontology
- seamless integration through use of Protégé 4/5 plugin mechanism
- based on DL-Learner Machine Learning framework to generate suggestions
- use of efficient and sophisticated machine learning algorithms
- can handle cases where an ontology is already damaged, i.e. adding the desired axiom renders the ontology inconsistent – in this case it displays the instances causing inconsistency
- Version 0.8 (February 2015):
- made plugin Protege 4.3/5.0 compliant
- Version 0.7 (April 2012):
- made plugin Protege 4.2 compliant
- source code improvements
- Version 0.6 (March 2010):
- fixed bug, which caused inconsistency not to be shown (suggestions were sometimes not highlighted in red even if they lead to an inconsistent ontology)
- allow fine-grained selection of language features to use (negation, quantifiers, cardinality restrictions)
- display minimum and maximum length of searched expressions (this allows to users to adapt the algorithm runtime to their needs and gives an impression of the complexity of searched expressions within the time limit)
- Version 0.5.2 (May 2009):
- fixed bugs #2790483 and #2791135, corrected manifest file, ensured Java 5 compatibility
- Version 0.5.1 (April 2009):
- corrected display of suggested class expressions; message when learning algorithm is still running
- Version 0.5 (April 2009):
- better integration in Protege (thanks to Nick Drummond for his support)
- switch from general machine learning algorithm to special purpose algorithm for definitions and super class axioms
- better performance through stochastic coverage testing procedures in machine learning algorithm
- visualisation of learned results in Protege to simplify the decision whether to add the axiom for the knowledge engineer
- Version 0.1 (December 2009): initial release
See also the DL-Learner page in the Protégé wiki.