Ontology: Definition of 'Quantity' has to be completed

[stap-m]

Description of the issue

There are two ideas for definitions given:

_A quantity is a property that is quantifiable by measurement._
_A number together with a unit of measurement to quantify an entity._

Ideas of solution

The ideas have to be discussed and decided on one.

Workflow checklist

• [ ] I discussed the issue with someone else than me before working on a solution
• [x ] I already read the latest version of the workflow for this repository
• [ x] The goal of this ontology is clear to me

I am aware that

• [ x] every entry in the ontology should have an annotation
• [ x] classes should arise from concepts rather than from words
• [x ] class or property names should follow the UpperCamelCase

[akleinau]

this should make sense with issue #12 Instances for units. It replaces our UnitOfMeasurement class (and subclasses) with the unit class (and subclasses) of the uo ontology.
A unit is defined there as "A unit of measurement is a standardized quantity of a physical quality."
So our definition of quantity should fit to that definition.

[akleinau]

https://github.com/OpenEnergyPlatform/ontology/issues/176#issuecomment-564946976
@fabianneuhaus also commented on that with:
"If I remember correctly, In the metrology world mass would be an instance of the class quantity. 71 kq would be an instance of the class quantity value.

But BFO has a different view on quantities (and, thus, indirectly on measurements) that is in some aspects philosophically sound, in others completely broken and sometimes just unclear. At some time in the future I would like to clean up that mess. But in the interest of making progress now, I think just going along with BFO probably makes sense."

[jannahastings]

BFO does not offer a definition for quantity and indeed explicitly avoids doing so. Mass is a quality in BFO. Other entitities that can be quantified may have quite different types.

IAO also avoids including 'quantity' and instead offers the 'measurement datum' branch. For example, 'mass measurement datum' is defined as: 'A scalar measurement datum that is the result of measurement of mass quality.' However, it is clear that what they mean is in line with the proposed definition of quantity as 'A number together with a unit of measurement to quantify an entity.' as they define the immediate parent as 'scalar measurement datum' and only numbers can be scalar. Moreover, they use two additional object properties to axiomatise the class as follows: '(has measurement unit label only mass unit) and (is quality measurement of only mass)' This captures the sense that the number has a unit -- in this case a mass unit -- and is about a particular entity -- in this case mass.

[akleinau]

If BFO avoids this then I like the idea of splitting it into quantities and quantity values.
The quantity value pair would then get the "A number together with a unit of measurement to quantify an entity." definition and be in line with IAO.
We'd just need classification (right now generically dependent continuant?) and a def for quantity

[akleinau]

so definition proposal:
unit of measurement (can't be changed bc imported): "A unit of measurement is a standardized quantity of a physical quality." (should get subclass of quantity?)
quantity: "A quantity is an immaterial entity defining an amount that can be measured."
quantity value: "A quantity value is an immaterial entity defined by a number together with a unit of measurement to quantify an entity."

[stap-m]

Just for my understanding: Do you mean a 'quantity' is composed of a 'quantity value' and a 'unit of measurement'? Would 'quantity value' and 'unit of measurement' be a subclasses of 'quantity'?

As I understand #220, there are no material/immaterial entities anymore in the new bfo version? We'd need a different classification of quantity/quantity value. We could make 'quantity' a 'quality' like 'mass quality' in IAO?

[akleinau]

The result of #220 is that we remain with our BFO 2.0 version because the proposed one there is an actualised version of the old BFO 1.0 version. So material/ immaterial remain in our ontology.

Rather a quantity value is composed of a unit of measurement/ quantity and a value.
A Unit of measurement is a standardized quantity. Maybe we should delete quantity? Unit of measurement and quantity value should be enough.

[akleinau]

I found a class numeral in our ontology. So we can make:

quantity value: "A quantity value is an immaterial entity defined by a numeral together with a unit of measurement to quantify an entity."

using the imported classes:
unit of measurement : "A unit of measurement is a standardized quantity of a physical quality."
numeral: "A symbol that denotes a number."

[akleinau]

@fabianneuhaus would that be correct?

[fabianneuhaus]

yes and no. :-) It may be the best solution available right now. Let's talk about it tomorrow.

[akleinau]

Ok I'll implement:
quantity value: "A quantity value is an immaterial entity defined by a number together with a unit of measurement to quantify an entity."
as discussed yesterday in Magdeburg.
When there's time we have to discuss how to implement numbers, but for now this should be enough.

[jannahastings]

See also: https://github.com/obi-ontology/obi/wiki/Data-and-Values

[akleinau]

so the idea where we left of was:
change quantity to quantity value:
quantity value: "A quantity value is an immaterial entity defined by a numeral together with a unit of measurement to quantify an entity."

use for the rest the already imported classes:
unit of measurement : "A unit of measurement is a standardized quantity of a physical quality."
numeral: "A symbol that denotes a number."

But I'm thinking now, maybe power is actually neither a quantity value nor a unit of measurement and we need the quantity class?

[akleinau]

maybe a "quantity concept" class (name can be discussed) would actually be helpful.
Then we could have power as a quantity concept which is in some way related or instanciated or implemented by the "power unit" unit of measurement.

I'm reluctant to just call that class "quantity" because the "unit of measurement" class refers to quantity in it's def and I''m not sure if that would be consistent with our idea of quantity?

[jannahastings]

I actually really don't think we should classify power as a quantity. (Really sorry I am coming to this late and the new power definition (#79) already went in last week.)

The main reason for this is, I don't think having quantity as a parent makes an accurate statement about what sort of thing power is. Quantities (= numbers) are abstract entities and as such they are not the sort of things that can have any direct effect in the world, while power is a physical thing that does have an effect in the world.

The secondary reason is that power can have different units of measurement (e.g. Watt vs. Horsepower). With each different unit of measurement you get a different number. So the number is not itself the thing that defines the power. Thus power is not a quantity.

I propose: We keep "quantity value" for this class, as it was earlier suggested (A number together with a unit of measurement quantify another entity) and make that a subclass of 'generically dependent continuant'.

Then we re-classify "power" (which will involve updating the definition).

I think that it is most natural to think of power as a process attribute, where the process is energy transformation. Process attributes are analogous to dependent continuants but for processes rather than continuants. While these are strictly not accommodated in BFO as dependent entities (as there is no second-order dependence), BFO does offer "process profile" as something _close_ to process attributes. For clearer terminology, I recommend we create our own class _process attribute_ as a subclass of _process profile_.

Then we define: Power is a _process attribute_ that is the derivative of energy transformation over time. (or: the rate of energy transformed over time?)

And we relate power to the class 'energy transformation' using a new object property 'process attribute of'.

[stap-m]

@jannahastings I fully agree.
Do we need further discussion on power? Then we should open a seperate issue. Else pull request.

[akleinau]

This seems to be consens of 4 people now so I'll implement.
I just need a def for process attribute and 'process attribute of'.

[akleinau]

Then we define: Power is a process attribute that is the derivative of energy transformation over time. (or: the rate of energy transformed over time?)

this is the comment, I changed the definition instead: "Power is the process attribute that represents the amount of energy transformed or transferred per time unit."

[jannahastings]

Attributes are rather than represent so:

Power: "Power is the process attribute that _is_ the amount of energy transformed or transferred per time unit."

I don't have a good definition for process attribute, but here are some candidate bad ones:

Process attribute: A process profile that describes a process as a whole.
process attribute of: A relation between a process and an attribute that describes that process

[akleinau]

I just heard these concepts for the first time, @fabianneuhaus can you help optimize these definitions of "process attribute" and "process attribute of"?

[akleinau]

@jannahastings is there an ontology or website where we can get good definitions?
Or do we just implement your candidate bad ones for now?

[jannahastings]

I don't think there are any good definitions, that's why this is not in BFO :-/.

An alternative would be that we just classify "power" beneath "process profile" directly, which would mean the definition would read as:

Power: "Power is the process profile that is the amount of energy transformed or transferred per time unit."

I asked @fabianneuhaus again if he has any opinions on this question, let's see.

[fabianneuhaus]

To me it seems that the best way to deal with the issue would be to break with BFO's definition of dependent particular: I would suggest that dependent continuant may depend on either independent continuants or occurrents. Hence, qualities may inhere in occurrents.

The reason is as follows: let's assume we extend BFO by a p-quality (we can call it "process attribute", but that's just a label that hides that we introduce qualities for processes). In this case we would have qualities like mass, length, surface and temperature as qualities. But speed, force, acceleration and rate of heat flow as p-qualities as completely separate kind of entities on a different branch of the ontology.

However, this is contrary to how physicists use these qualities.
E.g., in an equation like force = mass * acceleration physicists don't seem to treat qualities of independent continuants and qualities of processes in any way different. The same happens when they define units of measurements in equations like N = kg·m / s^2 .

If physicists don't make the distinction, we better have a good reason to make one. In particular, since it would lead to a lot of duplication. E.g., in addition to the distinction between Quantity and Base Quantity we would need p-Quantity and p-Base Quantity. And we wouldn't just need Quantity Value but p-Quantity Value. Measurement unit and p-Measurement unit. Etc, etc.

[jannahastings]

I am sympathetic to the sentiment of this proposal, but I object, for two reasons:

Pragmatically - we do use BFO and the corresponding relationships from RO, and introducing "process quality" as a subclass of quality that inheres in (depends on) processes will give inconsistency errors when used together with the axiomatization of BFO.

Technically - I do not think it is true that having "process attribute" distinct from quality means that we are forced to proliferate all the quantity classes and measurement unit classes. These are defined in the information section at the moment and are thus free to be about any kind of entity. In particular, there is no axiom that restricts measurement units to be about qualities in the BFO sense.

[fabianneuhaus]

No, but if you would try to define measurement unit in the BFO context, you would run into trouble with multiple inheritance if measurement units are measuring qualities and p-qualities.

Assuming the requirements are that we extend BFO and don't upset the apple cart by having p-qualities in the continuant branch, here is an alternative solution. We elucidate process attribute as follows:
A process attribute of a process P is an occurrent that existentially depends on P and
has a magnitude that can be expressed by the combination of (1) a scalar or a vector and a (2) a measurement unit, a measurement procedure, a reference material, or a combination of such.

I wouldn't call it a definition, because it isn't. And I can only stress that I think that this approach is much inferior to my previous suggestion. But the latter part of the elucidation is based on the International Vocabulary of Metrology, who should know what they are talking about.

[akleinau]

so we have the choice between breaking BFO's definitions whilst not confusing physicists or having an inferior solution with problems but without obvious breakage that is in line with metrologists

[akleinau]

But we are sure we need a class "power" and "power unit" is not enough?

[l-emele]

I do not fully understand the discussion, but I want to share an aspect which I think is important

There are quantities which have more than one units, like energy.
Example: An electric energy quantity of 1000 kWh (kilowatthour) is typically used by a single person household. That identical energy quantity can be expressed as 3600 MJ (megajoule)..

So you can apply different energy units without changing anything to the energy quantity itself. (There are many more energy units than only kilowatthour and megajoule.)

[jannahastings]

I recommend we use:

Power: "Power is the process attribute that is the amount of energy transformed or transferred per time unit."

Process attribute: A process attribute is a dependent occurrent that existentially depends on a process.

The parent of process attribute would then be occurrent.

I believe this is in line with the sentiment of @fabianneuhaus's proposal, but avoids using the mathematical terminology (magnitude, scalar, vector etc) which would belong better in the information entity branch. We are trying to classify the physical entity _power_, not the measurement datum with a specific value and unit.

[stap-m]

@akleinau will this PR be solved until the fist release?

[akleinau]

@akleinau will this PR be solved until the fist release?

Yes, it's ready to be implemented if we agree on @jannahastings definitions

[stap-m]

@jannahastings proposal sounds good to me. Even though it is not very intuitve :D
Do we need a new relation has process attribute such that energy transformation has process attribute some power or else?

[akleinau]

@jannahastings proposal sounds good to me. Even though it is not very intuitve :D
Do we need a new relation _has process attribute_ such that energy transformation has process attribute some power or else?

good idea!