Modular Structures and Atomic Decomposition in Ontologies
Authors: Chiara Del Vescovo, Matthew Horridge, Bijan Parsia, Uli Sattler, Thomas Schneider, Haoruo Zhao
JAIR 2020 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this section, we complement the conceptual and theoretical contributions from the previous sections with an empirical study. For this purpose, we use an implementation of the decomposition algorithm described in Algorithm 1 using syntactic locality-based modules ( , , and ), run it on a suitable corpus of ontologies, and evaluate the feasibility of this algorithm as well as the structure and granularity of the resulting ADs. |
| Researcher Affiliation | Collaboration | Chiara Del Vescovo is affiliated with British Broadcasting Corporation (industry). Matthew Horridge is affiliated with Stanford University (academia). Bijan Parsia, Uli Sattler, and Haoruo Zhao are affiliated with University of Manchester (academia). Thomas Schneider is affiliated with Universität Bremen (academia). Since there is a mix of industry and academic affiliations, the paper is classified as a collaboration. |
| Pseudocode | Yes | In Algorithm 1 we give the pseudocode of an algorithm for the computation of the AD that interleaves the above two points by using so-called key axioms for which to record modules, and by ensuring that, for each genuine module/each atom, exactly one key axiom is identified. |
| Open Source Code | No | The paper states: "We evaluate the implementation of Algorithm 1 that is part of the OWL API, namely the one available via Maven Central (maven.org) with an artifact Id of owlapi-tools." This indicates the authors *used* an existing open-source tool for their evaluation, rather than providing their own specific source code for the methodology described in their paper. |
| Open Datasets | Yes | To build our corpus, we started with the snapshot of the NCBO Bio Portal ontology repository18 by Matentzoglu and Parsia (2017), which contains 438 ontologies. [Footnote 18]: https://bioportal.bioontology.org/ |
| Dataset Splits | No | The paper describes how the corpus was curated (removing ABox axioms, empty ontologies, and non-OWL 2 DL ontologies) and analyzes the characteristics of this corpus (size, length, OWL profiles). However, it does not mention any training, test, or validation splits, as the experiments evaluate the performance of an algorithm on a set of ontologies rather than training a machine learning model requiring such splits. |
| Hardware Specification | Yes | All experiments have been performed on Intel(R) Core(TM) i7-6700HQ CPU 2.60GHz RAM 8GB, running Java version 1.8 JDK with an initial heap size of 1 GB and a maximal heap size of 8 GB. Time is measured in CPU time. |
| Software Dependencies | Yes | We use OWL API Version 5.1.11 to compute -, -, and -AD. All experiments have been performed on Intel(R) Core(TM) i7-6700HQ CPU 2.60GHz RAM 8GB, running Java version 1.8 JDK with an initial heap size of 1 GB and a maximal heap size of 8 GB. |
| Experiment Setup | Yes | We set 6 hours as the timeout for computing each kind of AD, i.e., we used a total timeout of 18 hours for all three kinds of AD. If the ontology cannot be decomposed within 6 hours for at least one kind of AD, we say that this ontology could not be decomposed due to a timeout. ... Time is measured in CPU time. |