Wikidata:Property proposal/unit of measurement

From Wikidata
Jump to navigation Jump to search

unit of measurement[edit]

Return to Wikidata:Property proposal/Natural science

   Under discussion
Descriptionunit in which a quantity is measured
Representsunit of measurement (Q47574)
Data typeItem
Domainsubclasses of individual quantity (Q71550118) or its subclasses
Allowed valuesinstances of unit of measurement (Q47574) or its subclasses
Example 1mass (Q11423)kilogram (Q11570)
Example 2area (Q11500)square metre (Q25343)
Example 3speed (Q3711325)metre per second (Q182429), kilometre per hour (Q180154)
Example 4magnetic moment (Q242657)ampere square metre (Q71581529)
SourceISO/IEC 80000 (Q568496)
Planned usemove unit-valued statements from measured by (P1880) to this new property; enter all values listed in ISO/IEC 80000 (Q568496); check consistency with the Wolfram Language (Q15241057) quantity and unit ontology (using the function QuantityVariableCanonicalUnit[...] and CompatibleUnitQ[...])
Number of IDs in sourcehundreds
Expected completenessalways incomplete (Q21873886)
See alsomeasured physical quantity (P111) (inverse of this new property), measures (P2575) (from measurement device to quantity), measured by (P1880) (from phenomenon to scale)

Motivation[edit]

The property measured by (P1880) is currently used in some hybrid way: It links a) a phenomenon (earthquake) to a scale (Richter scale) and b) a quantity (length) to a unit (metre). I find those uses are disjoint and could better be handled by two properties, each with their separate set of constraints (P1880 accepts as values "scales", and the proposed property "units").

Summary of the history of the existing measured by (P1880) (as I understand it): In the proposal discussion the description mentions units, but the examples are all of type a). Later the property was used mostly for type a), with a few uses of b). Some month ago I started working intensively on improving the quantity and unit ontology in Wikidata, and - lacking an alternative - continued entering units for quantities (use b)). I now feel that this mix-up of concepts will make the use, maintenance, choice of label and description, and consistency checks more difficult.

Therefore I propose this new property. Toni 001 (talk) 14:51, 22 December 2019 (UTC)

There is precedent for modeling quantities and units by making a connection from the quantity to the unit (not all provide the other direction). For instance:

Toni 001 (talk) 09:26, 8 January 2020 (UTC)

Discussion[edit]

Pictogram voting comment.svg Comment If this is a strict inverse of the existing property then it's really not needed I think. ArthurPSmith (talk) 19:31, 23 December 2019 (UTC)
In principle measured physical quantity (P111) should be enough to state all quantity-unit connections. However, ISO 80000-1:2009: Quantities and units—Part 1: General (Q26711930) emphasizes that "quantities come first and units then follow". If we want to follow this interpretation then it seems natural to have unit statements on the quantity items. But then there is also a more practical issue to consider: There might be tens or hundreds of quantities that are measured in, say, metre (Q11573): With the currently available P111 the metre items will become very crowded. This is somewhat analogous to listing all cities on the country items. The problem is probably the worst for 1 (Q199) (the "dimensionless unit"). Toni 001 (talk) 12:51, 24 December 2019 (UTC)
Symbol oppose vote.svg Oppose per Arthur (I fear a lot of bloat coming from this). Mahir256 (talk) 20:02, 23 December 2019 (UTC)
@Mahir256: Arthur raised some valid concern, so I gave some additional motivation why this property is useful and better suited than measured physical quantity (P111). Note that this is not simply creating additional data, but would be a better location for data already in measured by (P1880). For instance, now we can't (easily or cleanly) add property constraints that would tell an editor "hey, that's not a unit". And I've been reluctant to use P111 as that would mean adding countless statements of the form 1 (Q199) measured physical quantity (P111) static friction factor, 1 (Q199) measured physical quantity (P111) dynamic friction factor, 1 (Q199) measured physical quantity (P111) mass ratio, 1 (Q199) measured physical quantity (P111) linear expansion coefficient, 1 (Q199) measured physical quantity (P111) cubic expansion coefficient, 1 (Q199) measured physical quantity (P111) level of field quantity, ... to 1 (Q199) Toni 001 (talk) 10:09, 26 December 2019 (UTC)
Pictogram voting comment.svg Comment How about
pattern
quantity has dimension dimension
dimension unit of measurement unit / system of units system of units

example
speed (Q3711325) has dimension length per time
length per time unit of measurement centimetre per second (Q18413919) / system of units centimetre–gram–second system of units (Q26240)
example
coefficient of friction (Q1932524) has dimension dimension 1
dimension 1 unit of measurement dimensionless unit (Q1618549)

? This might help reduce bloat since there are fewer quantity–dimension associations and dimension–unit associations than there are quantity–unit associations.
And where there’s no concept of dimension, maybe
pattern
phenomenon “quantified on [scale]” (measured by (P1880)) scale

? (@Toni 001: It seems that maybe the vague label “measured by” was what invited in the first place the questionable ‘use b)‘ of measured by (P1880) which this proposal is now supposed to sort out. So a clearer might help also.)
BlaueBlüte (talk) 07:46, 8 January 2020 (UTC)
@BlaueBlüte: About your second point first: Yes, both label and description are (or were) vague, and looking at the property proposal discussion did not help because unit is mentioned there. Sometimes I have seen confusion about what the difference is between a unit and a scale, and this might have contributed to the confusing use of the property (P1880). Quick explanation: Temperature is measured by a temperature scale (say, Celsius scale, Kelvin scale, ...) and measure in the unit degree Celsius, Kelvin, ...; the scale is used to define the unit.
Now about your first point: torque (Q48103) and energy (Q11379) have the same ISQ dimension (P4020), yet their values are expressed in different units: newton metre (Q215571) and joule (Q25269), respectively (note that this is in the same unit and quantity system). Therefore the unit needs to be attached to the quantity. Toni 001 (talk) 08:14, 8 January 2020 (UTC)
@Toni 001: Good point about newton metre (Q215571) vs. joule (Q25269).
I just saw that there’s also unit of torque (Q21813488) and unit of energy (Q2916980)—and unit of length (Q1978718), since you brought up metre (Q11573) above. Can we fit those into this in any helpful way? E.g.,
pattern
quantity class of units of measurement unit of quantity
unit instance of (P31) unit of quantity
Then we’d have rather than (speaking of crowding)
example
length (Q36253) unit of measurement metre (Q11573)
length (Q36253) unit of measurement yard (Q482798)
length (Q36253) unit of measurement foot (Q3710)
width (Q35059) unit of measurement metre (Q11573)
width (Q35059) unit of measurement yard (Q482798)
width (Q35059) unit of measurement foot (Q3710)
depth (Q3250078) unit of measurement metre (Q11573)
depth (Q3250078) unit of measurement yard (Q482798)
depth (Q3250078) unit of measurement foot (Q3710)
depth (Q930412) unit of measurement metre (Q11573)
depth (Q930412) unit of measurement yard (Q482798)
depth (Q930412) unit of measurement foot (Q3710)

As for “measured by”, the more specific preposition on might help clarify the semantics of measured by (P1880) and be sufficiently common, cf. some usage statistics.
BlaueBlüte (talk) 09:22, 8 January 2020 (UTC)
@BlaueBlüte: A general remark: Classes like unit of length (Q1978718) are (very closely related to) what is called kind of quantity, a concept which is explained nicely in Defining 'kind of quantity' (Q71548419). The idea is that length, radius, wavelength, ... are all kind of length and lengths are measured in metre, millimetre, inch, .... This is already modeled to some extend, for instance by statements like radius (Q173817) subclass of (P279) length (Q36253) and metre (Q11573) measured physical quantity (P111) length (Q36253). Though this has not been stated anywhere formally, yet (to my knowledge). Considering length (Q36253) as both a quantity and a kind of quantity makes classes like unit of length (Q1978718) unnecessary. Tackling kind of quantity is on my to do / wish list: By proposing one or more models (either with or without new properties) and forming consensus.
Now to this property: You bring up a good point about unit of measurement-statements becoming crowded. The way out is to not list all possible units, but only typically used units or units given in references. So, while length - by being a very general term - might have quite a few statements, atomic mass (Q3840065) might only list kilogram and dalton.
I have updated the motivation with a (non-exhaustive) list of well-known ontologies that model the quantity-unit relation in the same way as proposed here. Toni 001 (talk) 10:16, 8 January 2020 (UTC)