In the WG meeting of 13 May 2016, a suggestion was made for a new property to specify the length of a time series, if a dataset includes a time series.
We are soliciting feedback from the community on such a new property.
The questions are as usual:
1. Is the information for this property available in existing statistical systems and applications?
2. How will exposing this information to general data portals enhance the discoverability of statistical datasets?
3. Do you know of any property in existing RDF vocabularies that could hold this information?
Component
DocumentationCategory
improvement
Login or create an account to comment.
Comments
Makx, can we please confirm that by length we are referring to the number of observed values in a time series?
If the intention is to specify when the first observation is made and the last, then Dublin Core already has a dcterms:temporal property. The property is supposed to point to another instance of the type PeriodOfTime. Unfortunately, that type is very loosely defined.
The property dct:temporal is already available in DCAT-AP as an optional property on Dataset. For its range, dct:PeriodOfTime, DCAT-AP recommends the use of schema:start and schema:end.
The property dct:temporal could be used to express the start and end of the time series. This is different, semantically, from the 'length' of the series but it does convey the same information.
Using dct:temporal would make the addition of a separate 'length' paramente unnecessary.
I'm unfamiliar with schema:start and schema:end. Do you have a link to the specification?
I think that the dct:temporal property with schema:start and schema:end meets the needs, and sounds easier to keep up to date than a 'length' property.
The properties schema:start and schema:end do not exists. My intention was to refer to schema:startDate and schema:endDate.
@Makx
Would that cover it? There is also schema:startTime and schema:endTime, should we need more precision (e.g. for a time series starting and ending on the same date).
@Uroš
It seems like schema.org itself has created confusion on this point. The definition of schema:startDate says "start date and time" and is used for example on schema:Event, while schema:startTime notes that "Event uses startDate/endDate instead of startTime/endTime, even when describing dates with times. This situation may be clarified in future revisions."
DCAT-AP uses PeriodOfTime with schema:startDate and endDate, so it will be more coherent to do the same in StatDCAT-AP. If at some point in the future, the expected clarification at schema.org recommends the use of startTime over startDate, we'll change all occurrences in all our Application Profiles.
On second thought, the domain for both properties is expected to be one of the following: CreativeWorkSeason, CreativeWorkSeries, DatedMoneySpecification, Event, Role. Although one could expect the definition of Event to be close to that of PeriodOfTime, the class description says otherwise:
"An event happening at a certain time and location, such as a concert, lecture, or festival. Ticketing information may be added via the 'offers' property. Repeated events may be structured as separate Event objects."
I don't think we need to be in line with DCAT-AP here, as it seems the semantics of the properties there were misinterpreted.
@Uroš
The requirement to be in line with DCAT-AP is a strong one. I would very much recommend not to create a incoherence between DCAT-AP and StatDCAT-AP.
Apart from that, three points that we may want to consider:
Proposed resolution: No extension necessary. The existing DCAT-AP property dct:temporal can be used.
The requirement to provide information of the length of time series can be met through the use of the property dct:temporal which is already available in DCAT-AP. The values of the property are members of the class dct:PeriodOfTime, with properties schema:startDate and schema:endDate.
@Makx,
Thanks. I have to say stress that a schema:Event is still "an event happening at a certain time and location," (such as a concert), and the domain is still part of the property semantics.
Having said that, my assumption wasn't based only on the HTML specification, but also on the RDFa one, and upon closer inspection of the latter, I discovered that the property domains are not declared via rdfs:domain, but schema:domainIncludes (a similar property is used for ranges), which definitely makes the semantics a lot more flexible.