UCT CS Research Document Archive

The TDDonto Tool for Test-Driven Development of DL Knowledge bases

Lawrynowicz, Agnieszka and C. Maria Keet (2016) The TDDonto Tool for Test-Driven Development of DL Knowledge bases. In Lenzerini, Maurizio and Rafael Penaloza, Eds. Proceedings 29th International Workshop on Description Logics (DL'16) 1577, Cape Twon, South Africa.

Full text available as:
PDF - Requires Adobe Acrobat Reader or other PDF viewer.

Abstract

Adding knowledge to an ontology lacks a verification step by the modeller in most cases, other than `try and see what the reasoner says about it'. This is due to the lack of a systematic testbed for ontology authoring. Reusing the notion of {\em Test-Driven Development} (TDD) from software engineering for ontology development resulted in the specification of 42 test types for the $\mathcal{SROIQ}$ language features, as TBox tests using its axioms and as ABox-driven tests with explicitly introduced individuals.
We developed TDDOnto, which implements that subset of the TDD tests that could be done by leveraging extant technologies. We examined what the most efficient implementation strategy is with 82 ontologies.
The TBox SPARQL queries with OWL-BGP were faster than the ABox-based approach except for disjointness, that effect is more pronounced with larger ontologies, and the OWL API approach is faster than the SPARQL queries for OWL 1 ontologies. A significant difference in performance between OWL and OWL 2 DL ontologies was observed. On average, the TDD tests are faster than classification reasoning, indicating that TDD tests are a promising alternative to the `try and see' approach in ontology authoring.

EPrint Type:Conference Paper
Keywords:Ontologies, OWL, Test-Driven development
Subjects:I Computing Methodologies: I.2 ARTIFICIAL INTELLIGENCE
ID Code:1085
Deposited By:Keet, C. Maria
Deposited On:18 August 2016
Alternative Locations:http://ceur-ws.org/Vol-1577/, http://www.meteck.org/files/TDDtestsDL16.pdf