Class expression learning for ontology engineering
Section snippets
Introduction and motivation
The Semantic Web has recently seen a rise in the availability and usage of knowledge bases, as can be observed within the Linking Open Data Initiative, the TONES and Protégé ontology repositories, or the Watson search engine. Despite this growth, there is still a lack of knowledge bases that consist of sophisticated schema information and instance data adhering to this schema. Several knowledge bases, e.g. in the life sciences, only consist of schema information, while others are, to a large
Preliminaries
For an introduction to OWL and description logics, we refer to [4] and [17].
Finding a suitable heuristic
A heuristic measures how well a given class expression fits a learning problem and is used to guide the search in a learning process. To define a suitable heuristic, we first need to address the question of how to measure the accuracy of a class expression. We introduce several heuristics, which can be used for CELOE and later evaluate them.
We cannot simply use supervised learning from examples right-away, since we do not have positive and negative examples available. We can try to tackle this
Efficient heuristic computation
Most of the runtime of a learning algorithm is spent for computing heuristic values, since they require expensive reasoner requests. In particular, retrieval operations are often required. Performing an instance retrieval can be very expensive for large knowledge bases. Depending on the ontology schema, this may require instance checks for many or even all objects in the knowledge base. Furthermore, a machine learning algorithm easily needs to compute the score for thousands of expressions due
Adaptation of the learning algorithm
While the major change compared to other supervised learning algorithms for OWL is the previously described heuristic, we also made further modifications. The goal of those changes is to adapt the learning algorithm to the ontology engineering scenario: for example, the algorithm was modified in order to introduce a strong bias towards short class expressions. This means that the algorithm is less likely to produce long class expressions, but is almost guaranteed to find any suitable short
The Protégé plugin
After implementing and testing the described learning algorithm, we integrated it into Protégé and OntoWiki. Together with the Protégé developers, we extended the Protégé 4 plugin mechanism to be able to seamlessly integrate the DL-Learner plugin as an additional method to create class expressions. This means that the knowledge engineer can use the algorithm exactly where it is needed without any additional configuration steps. The plugin has also become part of the official Protégé 4
The OntoWiki plugin
Analogous to Protégé, we created a similar plugin for OntoWiki [3]. OntoWiki is a lightweight ontology editor, which allows distributed and collaborative editing of knowledge bases. It focuses on wiki-like, simple and intuitive authoring of semantic content, e.g. through inline editing of RDF content, and provides different views on instance data.
Recently, a fine-grained plugin mechanism and extensions architecture was added to OntoWiki. The DL-Learner plugin is technically realised by
Evaluation
To evaluate the suggestions made by our learning algorithm, we tested it on a variety of real world ontologies of different sizes and domains. Please note that we intentionally do not perform an evaluation of the machine learning technique as such on existing benchmarks, since we build on the base algorithm already evaluated in detail in [26]. It was shown that this algorithm is superior to other supervised learning algorithms for OWL and at least competitive with the state of the art in ILP.
Related work
Related work can be categorised into two areas: The first one being supervised machine learning in OWL/DLs and the second being work on (semi-)automatic ontology engineering methods.
Early work on supervised learning in description logics goes back to e.g. [9], [10], which used so-called least common subsumers to solve the learning problem (a modified variant of the problem defined in this article). Later, [7] invented a refinement operator for and proposed to solve the problem by using a
Conclusions and future work
We presented the CELOE learning method specifically designed for extending OWL ontologies. Five heuristics were implemented and analysed in conjunction with CELOE along with several performance improvements. A method for approximating heuristic values has been introduced, which is useful beyond the ontology engineering scenario to solve the challenge of dealing with a large number of examples in ILP [35]. Furthermore, we biased the algorithm towards short solutions and implemented optimisations
References (35)
- et al.
Computing the least common subsumer w.r.t. a background terminology
Journal of Applied Logic
(2007) An Introduction to Categorical Data Analysis
(1997)- et al.
Approximate is better than “exact” for interval estimation of binomial proportions
The American Statistician
(1998) - S. Auer, S. Dietzold, T. Riechert, Ontowiki – A Tool for Social, Semantic Collaboration, in: ISWC 2006, vol. 4273 of...
- et al.
Completing description logic knowledge bases using formal concept analysis
- et al.
A refinement operator for description logics
- et al.
Occam’s razor
- et al.
Computing least common subsumers in description logics
- et al.
Learning the CLASSIC description logic: theoretical and experimental results