Today, we released DL-Learner Build 2008-10-13. DL-Learner is a tool for learning complex class descriptions from examples and background knowledge. It extends Inductive Logic Programming to Description Logics and the Semantic Web.
Downloads are available at the sourceforge.net project page. For a list of the most important changes since the previous release (Build 2008-02-18), see the ChangeLog. Some notable features are:
- addition of a new learning algorithm, which uses background knowledge more efficiently to find solutions of learning problems
- a GUI as interface to create or modify configuration files and execute algorithms
- a fast approximate instance checking algorithm decreasing the time for example coverage checks (the most expensive operation) significantly, thereby improving overall performance
- a matured fragment extraction algorithm, which allows to grab OWL-DL fragments from large knowledge bases (using SPARQL), which contain enough relevant information to conduct concept learning, while they are small enough to reason efficiently (more information)
DL-Learner can be used to:
- solve general supervised Machine Learning problems using ontologies as background knowledge (given as OWL files, SPARQL endpoints, etc.) , e.g. it was used to predict whether chemicals can cause cancer
- help knowledge engineers by learning definitions and subclass axioms (plugins for Protégé and OntoWiki in progress)
- support searching/navigating/recommendations in knowledge bases
