-
Notifications
You must be signed in to change notification settings - Fork 2
Shared Dimensions and Hierarchies
Similar to the concept tables Shared Dimensions are already existing concepts which can be used to describe data cubes (i.e. Cantons)
Under the tab Shared Dimensions
a new one can be created. There are currently two ways of starting a new Shared Dimension via Cube Creator:
- Define the Shared Dimension manually from scratch
- Import a
.trig File
from a exported dimension
Dimension name: The Shared Dimension needs a name which should be specific to the concept it represents Dimension identifier: You need provide a unique name for the Shared Dimension, which will internally used as identifier and used inside the published data.
By pressing the 🖊️ icon next to the dimension title the metadata of the Shared Dimension can be accessed. The fields have descriptions to give you hints on the content. Mind that you can also open other Shared Dimensions to compare if you are unsure about the usage of the fields.
Valid from / to:
Under the tab Dimension
you can find these two options which let you define a time frame of validity for the Shared Dimension. Especially the valid to
option is crucial for marking now invalid Shared Dimensinos as deprecated
Term Properties:
Under the tab Term Properties
you can define one or more predicates which will be used in relation to the objects in your Shared Dimension.
Name: Will be displayed when choosing relations in the Shared Dimension.
Required: Check this if the predicate is required for each of the entries in the Shared Dimension
Allow Multiple: Check this if this predicate can be used multiple time for the same entry
Value type: There are three different value types
1. Literal: Relationship is a specific data type (i.e. 'integer')
2. Shared Term: Relationship is to another Shared Dimension or within itself
3. Lang String: Relationship is based on a language
By pressing the + Add term
button, the menu will open. Each new term needs a Name which will be displayed in the given languages and an Identifier * that will be added to the URI and need to be unique for each new term. On the contrary the Identifiers are the indices used in a dataset for the term (i.e. the identifier '1' is commonly used for the canton Zurich).
Under the tab Dynamic Properties
the relationship of the term via the previously defined predicates can be determined. If there are multiple predicates via term properties defined they will all show up in this menu.
Shared dimensions can use terms defined in other Shared Dimensions and combine them with their own relationships.
To do this in the menu Term properties
the corresponting relationships with each of the shared dimensions need to be defined (i.e. in this example there are two properties. One within itself and the other to another Shared Dimension).
Defining and adding the Dynamic Properties works just as describe above.
A Hierarchy is actually a description
of a hierarchy that is based on existing links between terms of a same Shared Dimension or from one Shared Dimension to another (linking cantons to municipalities for example).
To start a new Hierarchy press the + Create hierarchy
button and the menu will open.
The top of the window allows to define:
Name: Will define the name of the hierarchy and will be displayed in the list of hierarchies
Root dimension: From which Shared Dimension a hierarchy will be formed
Root: Which term in the chosen Shared Dimension is the starting point / top of the hierarchy. A hierarchy typically starts from an "All" root node, but if the hierarchy does not have a single root node, but many, they can be added one by one.
Then the levels of the hierarchy must be defined:
You have to define each level of the hierarchy, to the depths of the deeper "leaf" of that tree.
Type: This is an optional value, to choose a specific Class for the terms of that level.
Property: This is the important and mandatory value, by choosing in the drop down the predicate that link the terms in the the Shared Dimension. This must be set according to the existing property that links the terms, and its direction.
The existing links are typically described with skos:broader/skos:narrower
, or for locations schema:containsPlace/containedInPlace.
As a hierarchy is defined from the parent to the children, check the Inverse
box if the Shared Dimenson defines the links from the children to the parents. Typically, skos:narrower
and schema:containsPlace
are from parent to children (do not check Inverse
), skos:broader
and schema:containedInPlace
are from children to parent and need the Inverse
check.
Here is an example of an existing hierarchy, with the root node "All organisms":
The published hierarchies are then at disposal for the cubes edition: when linking a dimension to a shared dimension in the Cube Designer, the hierarchy can also be added.