Ontology antipatterns are structures that re ect ontology modeling problems because they lead to inconsistencies or to bad reasoning performance. Based on a collection of antipatterns coming from our experience in ontology engineering projects and bad modeling practices found in the literature, we propose to represent them as SPARQL queries and conduct an experiment to detect them in an ontology corpus obtained from the Watson ontology search portal.
I.2.4 [Computing Methodologies]: Arti cial IntelligenceKnowledge Representation Formalisms and Methods Experimentation Antipattern detection, SPARQL query, OWL ontology
The focus of this poster is to understand, in the context of a larger experiment, how often antipatterns appear in existing publicly available ontologies and how these antipatterns are normally clustered in them, so as to understand better their characteristics and their importance in terms Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee.
EKAW ’2010 Lisbon, Portugal Copyright 2010 ACM X-XXXXX-XX-X/XX/XX ...$10.00. of addressing them later in the context of ontology debugging tools and tasks inside methodologies and methods. To achieve this objective, we have selected a set of ontologies from those available in the Watson semantic search engine, we have transformed the selected antipatterns from our catalogue into sets of SPARQL queries or have transformed the original ontologies into a form where simpler SPARQL queries can be run to detect antipatterns, and have run those queries against the selected ontologies or their transformations, switching on and o inferences, so as to see whether there is any impact in the detection and posterior repair of the ontologies. 2.
In [
Detectable Logical AntiPatterns (DLAP). They represent errors that DL reasoners and debugging tools normally detect. Our experiment is focused on the detection of three examples of DLAP antipattern: AndIsOr (AIO), OnlynessIsLoneliness (OIL), UniversalExistence (UE).
Cognitive Logical AntiPatterns (CLAP). They represent possible modelling errors that may be due to a misunderstanding of the logical consequences of the used expression. We try to detect two of them: SynonymOrEquivalence (SOE), SumOfSom (SOS).
Guidelines (G). They represent complex expressions used in an ontology component de nition that are correct from the logical and cognitive points of view, but for which the ontology developer could have used other simpler alternatives or more accurate ones for encoding the same knowledge. We choose to detect only one antipattern of this type namely DisjointnessOfComplement (DOC). 3.
We have elaborated four di erent strategies in order to detect antipatterns in OWL ontologies by means of SPARQL queries, based on the usage of the PatOMat ontology pattern detection tool1. This tool is part of the PatOMat enables the processing of a set of SPARQL queries over a set of ontologies, producing a report in terms of numbers of patterns detected (queries) and details for each ontology.
Transforming antipatterns into SPARQL queries is not a trivial task. In some cases several SPARQL queries are needed to represent all the possible versions of an antipattern, and some of these transformations are di cult to generate.
Now we will show the di erent approaches that we have followed in order to detect antipatterns in the collection of ontologies that we have selected.
In this approach, we take into account that SPARQL query engines per se do not consider inferences that can be done with OWL ontologies. However, our assumption is that there will be cases where this is the only solution that can be applied due to the fact that in some cases inconsistent ontologies can lead to di culties in the detection of antipatterns due to some wrong inferences being done across the whole hierarchy. Hence we write in general a large number of queries for each antipattern, so that we embed in the syntax of those SPARQL queries some of the inferences that should be done in the OWL ontologies, which could be done with a reasoner.
When it is possible to use a reasoner, we materialise all the inferences that can be done by an OWL reasoner on the selected ontologies and then run SPARQL queries over the resulting ontologies.
Due to the uncertainty about some results from reasoners, to the complexity of creating a large number of SPARQL queries for an antipattern, and nally to the fact that di erent ontology developers may have di erent implementation styles, what leads to very different alternative implementations of the same sets of axioms, and makes it di cult to nd all the possibilities in which queries should be generated in order to identify antipatterns, we propose to follow a two step process where we harmonise the implementation style of the ontology using transformations before proceeding to execute the queries.
In this strategy, we have explored the behaviour of the SPARQL query checking method on the asserted ontology after transformation. 1The release used for this poster is at: http://eso.vse.cz/ ~svabo/patomat/detectionTool.zip 4. SP Trans+R: Apply transformations on the original ontologies in order to harmonize the axioms, and search in the materialisation of these harmonised ontologies. In this strategy, we transformed the ontologies in order to harmonise the implementation style. Then, we have explored the behaviour of the SPARQL query checking method on the materialised ontology (also after transformation). 4.
We have used the Watson API to retrieve publicly available ontologies that we could run experiments with, and we have always accessed these ontologies using the Watson cache. We searched for ontologies satisfying the following constraints: they should be implemented in OWL, they should have at least ve classes, they should be classi ed as inconsistent by Pellet reasoner. So the corpus is nally composed of 66 incoherent ontologies.
The table 1 shows a summary of the total numbers of antipatterns detected on the selected ontologies using each of the methods that have been proposed.
method of detection
SP
SP+R
SP Trans SP Trans+R 29 24382
UE 15 3268 The results analysis shows that although antipatterns in the collection occur rarely, they tend to appear many times in the same ontology, re ecting a bad modeling practice. 5.
In this poster, we present how antipatterns can be detected using di erent methods and we have tried to show which is the best detection process to be used for each type of antipattern. In many cases, these antipattern detection tools are very sensitive to the implementation style of the ontology developer.