Open HLWeil opened 5 months ago
The current plan is to use MediaObject
for data fragments and annotate them through the variableMeasured
property in the assay/study Dataset
object. Specifically, we plan the following:
variableMeasured
of type PropertyValue
.MediaObject
, referenced from its file object through hsaPart
.PropertyValue
through the about
property.PropertyValue
objects point back through propertyID
, although this is not a valid jsonld link.
A graph explaining this model and a corresponding profile for MediaObject
, Dataset
and PropertyValue
will be added to this repository soon.@floWetzels @HLWeil
Are all the questions from this issue solved? There is a datamap profile. Is that profile ready to be implemented or are there still things that need consideration?
If that profile is ready to be implemented, that would be great as that would unblock https://github.com/nfdi4plants/ARCtrl/issues/432
Data entities of the datamap file should be of type schema.org/PropertyValue. Explanation see https://github.com/ESIPFed/science-on-schema.org/blob/master/guides/Dataset.md#variables.
For general description of a data entity (which can also be a fragment of a file), we use the MediaObject type. As the datamap adds further description to exactly this kind of data entity, we suggest to represent the entities of the datamap as both PropertyValue and MediaObject.
By this, fragment selectors (data localization) would be standardized between reference from the process and the datamap. Also, it could be interpreted according to Science-On-Schema.Org.
For visual reference, consider:
Now two questions arise:
How to depict datamap fields in PropertyValue?
How to reference PropertyValue from datasets (Assay/Study)?
variableMeasured
property. This would require the data entities to be directly linked from the dataset, not making use of thehasPart
property.hasPart
to reference data files. In this case the dataset would reference the files and the files themselves could reference the data entitities (fragments), again usinghasPart
(-> CreativeWork). This would make linking down from folders to files to fragments very streamlined. But this might cause problems with cases like in BioImaging, where we might have fragments of folders, not files.