Efficient Extraction of EL-Ontology Deductive Modules

Authors: Hui Yang, Yue Ma, Nicole Bidoit

AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our experiments on real-world ontologies show that our pseudominimal modules are indeed minimal modules in almost all cases (98.9%), and computing pseudo-minimal modules is more efficient (99.79 times faster on average) than the stateof-the-art method Zoom for computing minimal modules. Also, our complete modules are more compact than modules, but their computation time remains comparable.
Researcher Affiliation Academia Hui Yang, Yue Ma, Nicole Bidoit LISN, CNRS, Universit e Paris-Saclay yang@lisn.fr, ma@lisn.fr, nicole.bidoit@lisn.fr
Pseudocode No The paper describes the algorithm using text and equations but does not provide a formal pseudocode block.
Open Source Code Yes We implemented a prototype For Mod3 of our algorithm in Python and evaluated it over three real-world ontologies: https://gitlab.lisn.upsaclay.fr/yang/formod
Open Datasets Yes We implemented a prototype For Mod3 of our algorithm in Python and evaluated it over three real-world ontologies: Snomed CT (versions Jan 2016 and Jan 2021) and NCI (version 16.03d)4. We denote them as sn16, sn21, nci, respectively. Here, sn16 and nci are two EL-terminologies containing 317891 and 165341 axioms respectively. And sn21 is an EL-ontology with 362638 axioms.
Dataset Splits No The paper does not explicitly detail training, validation, or test splits. It mentions running experiments over '2 sets ΣO n of 1000 randomly generated signatures', but not data splits for model training/validation/testing.
Hardware Specification Yes All the experiments run on a machine with an Intel Xeon Core 4 Duo CPU 2.50 GHz with 64 Gi B of RAM.
Software Dependencies No The paper mentions the prototype For Mod is implemented in Python but does not provide specific version numbers for Python or any other libraries/dependencies.
Experiment Setup No The paper states: "For each ontology O, we run the experiments over 2 sets ΣO n of 1000 randomly generated signatures, where each signature has n concepts (n {50, 100}) and 10 roles." It also mentions a "total run-time limit of 600s". However, it does not provide details on hyperparameters or specific model configuration settings beyond this.