Ontology-Mediated Query Answering: Harnessing Knowledge to Get More from Data

Authors: Meghyn Bienvenu

IJCAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental This short paper provides an overview of two recent lines of research to which the author has contributed. Section 2 is concerned with understanding the limits and possibilities of query rewriting, a key algorithmic technique for OMQA, while Section 3 tackles the issue of making OMQA robust to data inconsistencies. Experimental evaluation showed that the rewritings produced by such rewriting engines were often huge, making them difficult, or even impossible, to evaluate. Experimental results on real-world ontologies are very encouraging: the vast majority of AQs do possess FOrewritings, and the computed rewritings (represented as NDL programs) are typically quite small.
Researcher Affiliation Academia Meghyn Bienvenu CNRS, Universit e de Montpellier, INRIA
Pseudocode No The paper provides examples of axioms and queries, but no formal pseudocode or algorithm blocks.
Open Source Code No The paper mentions the 'CQAPri system' but does not state that the authors are releasing source code for their described work or provide a link to it.
Open Datasets Yes Each hospital will store its patient data in databases, using terms from a standardized medical ontology (or linking to these terms via mappings). Moreover, by utilizing a standardized ontology, patient data from different hospitals can be seamlessly integrated. For example, in the EU-funded Optique project, industrial partners Statoil and Siemens are adopting OMQA to make it possible for end users to formulate their queries over multiple complex data sources. Beyond their use in nextgeneration enterprise information systems, OMQA also holds much promise in the areas of medicine and the life sciences, where significant energy has already been spent in developing high-quality ontologies, the large-scale medical ontology SNOMED being the most prominent example.
Dataset Splits No The paper describes the use of data in illustrative examples and mentions experimental evaluation but does not provide specific details on training, validation, or test dataset splits for reproduction.
Hardware Specification No The paper does not provide any specific hardware details such as CPU, GPU, or memory used for experiments.
Software Dependencies No The paper mentions the 'CQAPri system' and 'SAT solver' but does not specify versions for these or any other ancillary software dependencies like programming languages or libraries.
Experiment Setup No The paper discusses algorithmic approaches and experimental findings but does not provide specific details about the experimental setup, such as hyperparameters or system-level training settings.