Dataset Profile - Data Model Review

[KellyVance]

Original Profile

Link to Powerpoint for review: Dataset Profile Please use Review Tab in the powerpoint file to add NEW issues as comments (not additional boxes). Check existing comments boxes to add to the discussion on existing issues before creating new issues.Please reply to this comment with subjects of issues raised on powerpoint to track review progress.

@GSQ-AI @Greenwoodmatthew @KellyVance

[Greenwoodmatthew]

6 Comments added to ppt file on theme, survey or obs, FOI, access, location and foaf

[dxwell]

Some comments in response to the review:

What is Dcat Theme? This is the subject/category of the dataset. This is the highest level classification of the data.

• See https://www.w3.org/TR/vocab-dcat-2/#Property:resource_theme
• What are the categories of data that GSQ publishes.
• We need to be future thinking for this vocab - e.g. if DES start putting data into the portal, then their dataset theme would be perhaps "environmental data".
• There is a draft vocab at https://vocabs.gsq.digital/vocabulary/gsq-dataset-theme that was created to allow the pilot to go ahead. However, this needs a lot of work.
• I recommend that we have high level themes in this vocab - e.g. geoscience data, publications, maps, reports, etc.
• The theme of "geoscience data", which is then further described in the metadata "earth science data category" vocab. BTW - earth science data category needs to be shown in the diagram.
• John Mc currently has this vocab - let's do some collaboration to help John with this.

What is a dataset? Is a survey a group of observations?

• The reason this relationship between the dataset class and the survey class is to show that the enduring dataset is the result of a one-off event (e.g. a survey). Ways to say this are:
  • A survey or observation generates a dataset
  • A dataset was generated by a survey or observation
• For instance, a seismic survey event generates a dataset consisting of SEG-D, SEG-Y, ops logs, navigation logs, photos, etc.
• A company may then reprocess the SEG-D (an observation event) creating a new dataset. However, the provenance of the data is still the original survey.
• For instance, the Hylogger observation event creates a dataset comprised of high-resolution imagery and reflectance spectra mineralogical data.
• Having this relationship allows a user to:
  a. Understand the provenance of any dataset
  b. See all of the datasets that were generated by a particular survey or observation
• For more info, see SOSA hasResult
• For more info, see SOSA isResultOf
• For more info, see PROV wasGeneratedBy

FOAF will hold the most important metadata for searching...

• Correct - this will contain the level 1 and level 2 metadata.
• This is not yet defined in a schema, it is expected that this metadata will be expressed in key->value pairs and different datasets will have different schemas. e.g.

  {
  "company" : "Acme Pty Ltd",
  "address" :  {
    "street" : "123 Second St",
    "city" : "Brisbane",
    "state" : "Queensland" 
  }
  }

• This will be an ongoing work. The focus has been on the level 0 (the primary metadata).
• If people are ready to start documenting the level 1 and 2, then go for it.
• FYI John and Luke are reviewing the secondary metadata that is currently in GEM. This metadata will go into a document as key-value pairs. This metadata will be searchable and able to be read by person or machine.

I think we need defined access rights, theme, location and other vocabs before we can comment of dataset profile? Make sure you read the full page at https://github.com/geological-survey-of-queensland/gsq-dataset-profile, not just the diagram.
This lists the vocabs for access rights and theme.
Hopefully, the other content in this response helps provide understanding.

How is location held? GeoJSON?

Location can be defined in multiple ways:

• CKAN needs GeoJSON - at least a bounding box, ideally a polygon in addition.
• The dataset can contain a shape file in SHP, MAT or other format.
• The dataset can specify a WFS where it offers value. We have been discussing creating WFS for the geochemical dataset.
• Location can be expressed as a centroid, point, line, polygon, complex polygon, 3D, or even a named place.

I still think this needs a better explanation of purpose (FeatureOfInterest).

• Yep, we need to do more work in communicating how FeatureOfInterest will be used for GSQ.
• The original word in the SOSA model was "target" - i.e. what was the target of the survey? Were you trying to find out more about the basin? A particular formation? The geology of the permit area?

[dr-shorthair]

Note that there is no class dcat:Theme. The range of dcat:theme is skos:Concept - i.e. a term from a controlled vocabulary. The acceptable controlled vocabularies are listed in the properties of the dcat:Catalog.

[dxwell]

Note that there is no class dcat:Theme. The range of dcat:theme is 'skos:Concept`- i.e. a term from a controlled vocabulary.

Good to see you hooked into the review @dr-shorthair The intention was to say that dcat:Theme is a property of the class and that property is a term from a controlled vocabulary. Is it the statement This is the highest level classification of the data. that is misleading, i.e. the use of the word classification?

[dr-shorthair]

For location, also see https://www.w3.org/TR/vocab-dcat-2/#spatial-properties - these could also be serialized in JSON-LD. See here: https://github.com/w3c/dxwg/blob/dcat-issue-971/dcat/examples/spatial-properties.jsonld

[dr-shorthair]

The characterization as a 'high-level classification' is fine. It is your hijacking the dcat: namespace that I object to. gsq:Theme rdfs:subClassOf skos:Concept . would be acceptable in a profile of DCAT.