You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
In the NMDC, we define classes to capture complex object types.
For example, the class quantity value allows us to capture both the numeric and units parts of measurement. For example, the depth a biosample can be represented using:
We also use a has raw value property to capture value in the original data source (e.g., 5 meter). Not sure if that property is appropriate for COB, though.
This very similar to @jamesaoverton proposal for directly linking information to entity, which I think went something like:
:james#1 :has-quality :height#1 .
:height#1 :has-quantity 180; :has-unit cm .
Depending on the need, the has-attribute-value can take different attribute value classes in its range: text value, timestamp value, etc.
A possible advantage I see (right now) in doing this, is that allows you to more easily specify the domain the data properties (e.g., the domain of has-numeric-value is quantity value). Whereas the domain of the general has-quantity data property may have to allow for many different types.
Is this level of indirection worth the cost? Is it over-modeling? Is it simpler to define a measurement datum that is about the entity? E.g.:
I do like the "measurement datum" which can have parts - value, unit, and optional aboutness. This fits well with modelling a system's different components from a cognitive or robotics perspective. One assembly line component can be for measuring diameters of things (outputing a measurement datum) but it doesn't know what entity it is measuring. Another more general component that knows what entities are on the assembly line can an add an 'is-about' reference to the measurement datum, etc. in an integrative fashion.
In the NMDC, we define classes to capture complex object types.
For example, the class quantity value allows us to capture both the numeric and units parts of measurement. For example, the depth a biosample can be represented using:
We also use a
has raw value
property to capture value in the original data source (e.g.,5 meter
). Not sure if that property is appropriate for COB, though.This very similar to @jamesaoverton proposal for directly linking information to entity, which I think went something like:
But adds one more level of indirection. E.g.:
Depending on the need, the
has-attribute-value
can take different attribute value classes in its range:text value
,timestamp value
, etc.A possible advantage I see (right now) in doing this, is that allows you to more easily specify the domain the data properties (e.g., the domain of
has-numeric-value
isquantity value
). Whereas the domain of the generalhas-quantity
data property may have to allow for many different types.Is this level of indirection worth the cost? Is it over-modeling? Is it simpler to define a
measurement datum
that is about the entity? E.g.:The text was updated successfully, but these errors were encountered: