-
Notifications
You must be signed in to change notification settings - Fork 4
Meeting Notes 11.01.2021
Present:
- Emanuele Ghedini
- Jesper Friis
- Casper Andersen
- Joana Morgado
- Francesca Bleken
- Rickard Armiento
- Saulius Grazulis
- Stuart Chalk
An introductory round of all participants. The following links are related to some of the current efforts of the participants' other projects:
- Stuart, SciData: https://stuchalk.github.io/scidata/
- OPTIMADE: https://www.optimade.org
- Rickard, Materials Design Domain:
Emanuele asks the central question: What is the goal of the domain-crystallography?
Francisca and Jesper: Their goal is be able to define a crystal / describe a crystal structure, sufficiently effecient for DFT calculations/simulations.
However, how should this be implemented in the EMMO? Should it be modular? What should go in top/middle, etc.?
Saulius: Reuse the IUCr CIF dictionaries.
Understand that crystallographical concepts are closely related to the experimental techniques, specifically (X-ray/neutron/electron) scattering.
IUCr CIF framework: https://www.iucr.org/resources/cif
Stuart: https://www.iucr.org/__data/iucr/cifdic_html/1/cif_core.dic/index.html
Rickard: After checking the domain-crystallography repository it seems most of the essential concepts are already in there (since they're taken from the CIF dictionaries). Look at the OPTIMADE specification of structures versus what CIF defines and try to determine what is best for the ontology and its relations.
Jesper agrees with Saulius' sentiment that the ontology is a good opportunity to integrate with experimental concepts. Emanuele: This is exactly the point of EMMO: To link between the real measurable object and the theory. As a development approach he suggests to start from a given CIF concept and integrate it into the ontology. Then tag on relations and integrate those concepts. All one by one.
Stuart: Concerning the eventual format used for ontology-tagged data: Implement in JSON-LD. Stuart provided an example of a JSON-LD file using SciData. Among other things, he showed how he utilized different sets of ontologies and namespacing in a single data file to define what concepts the data objects refer to. Development of the SciData ontology/(-ies) started from articles' data tables -> SQL db -> JSON-LD data structures -> ontology Now it's a symbiotic relationship between the last two parts/steps.
Emanuele: Suggests to use RDF and OWL, where the goal is to store both the data but also all the metadata for said data. Then one can later create a mapping between the current domain-crystallography and the SciData ontology.
Saulius: Compound is an individual chemical compound. Substance is a mixture of nominal compounds.
CIF matched with HDF-5? Stuart: Not so familiar with HDF-5.
IUPAC definitions in CDATA? Stuart: It doesn't seem to be there yet.
Emanuele: We want to build an ontology first, format second. Mirror concepts from Stuart's cif.owl, but then develop a mapping ontology for the two ontologies. Do not just encapsulate and include an external ontology. This is preferred as to respect authorship rights.
Saulius: Can mapping between ontologies be automated? If a model is good for one thing and not another, how would one describe this in the ontology? Jesper: Granularity is built into EMMO.
Philosophical versus practical approach to ontological concepts (classes) and indivuduals (instances). This was the final topic, which was not completely resolved, but was discussed with the example of transferring knowledge about a C atom from a crystallographical experimental point-of-view to a more physical theoretical point-of-view.