LeDA

by Johanna Völker

A tool for learning disjointness axioms.

Versions & Compatibility

This section shows available versions of LeDA and for each version its compatible application versions and necessary plugins.

Changelog

Functional Description

LeDA is a tool for the automatic generation of disjointness axioms based on machine learning classification. The classifier, that determines disjointness for any given pair of classes, is trained based on a gold standard of manually created disjointness axioms. Each axiom of the gold standard is represented by a pair of classes associated with a label - disjoint or not disjoint - and a vector of feature values. As in our earlier experiments, we used a variety of lexical and logical features, which we believe to provide a solid basis for learning disjointness. These features are used to build an overall classification model on whose basis the classifier can predict disjointness for previously unseen pairs of classes.

For performance and useability reasons the LeDA plugin works with the following reduced set of features. The full feature set as described in D3.8.1 is so far only available in the original version of LeDA (http://ontoware.org/projects/leda/).

• leda.features.LabelSimilarity_JaroWinkler
• leda.features.LabelSimilarity_Levenshtein
• leda.features.LabelSimilarity_QGrams
• leda.features.Ontology_Subsumption
• leda.features.Ontology_Similarity
• leda.features.Ontology_ObjectProperties
• leda.features.TaxonomicOverlap_Subclasses
• leda.features.TaxonomicOverlap_Instances

Installation

The installation of this plugin does not require specific configuration steps, that might demand for in depth explanations. It can be performed by using the standard plugin update mechanisms of the NeOn Toolkit.

User Documentation

More information with regards to this plugin can be found in NeOn D3.8.1 and on the LeDA homepage (http://ontoware.org/projects/leda/).

LeDA View

This screenshots shows the main view of LeDA. A training or classification process can be triggered by means of a context menu, that is accessible by clicking on an OWL ontology in the navigator view on the left ("Learn Disjointness"), or by selecting "Run LeDA" from the main "LeDA" menu.

Preferences

The preference page is accessible from the main menu of Eclipse ("Window" -> "Preferences..." -> "LeDA Preferences"). It allows for setting a variety of parameters for both training and classification:

• Training or classification
• Classifier: The Weka (http://www.cs.waikato.ac.nz/~ml/weka/) classifier to be used by LeDA.

• Training
  • Training ARFF: This ARFF file will be generated automatically at the end of the training phase. At this point, it will contain all the necessary information for creating a classification model.
  • Training file: This is an optional, but recommended parameter. If no training file is specified, positive and negative examples will need to be generated from the training ontology - a process, that is very time-consuming.
  • Training ontology: This has to be an ontology, which contains a complete set of manually created disjointness axioms.

• Classification
  • Training ARFF: The training ARFF file must have been generated in a preceding training phase.
  • Classification ARFF: This file will be generated automatically as soon as the classification phase is finished and mainly serves debugging purposes.
  • Input ontology: This parameter specifies the ontology, which is to be enriched by disjointness axioms.

Category: Plugin