- From: Martin Hepp <mfhepp@gmail.com>
- Date: Tue, 12 Jun 2018 10:06:38 +0200
- To: Simon.Cox@csiro.au
- Cc: anthony.moretti@gmail.com, danbri@google.com, pfpschneider@gmail.com, richard.wallis@dataliberate.com, public-schemaorg@w3.org
One more point: In a huge ecosystem like that for schema.org, the overall quality of data can sometimes increase if we use a class or property name that is more broadly understood, even if technically less correct. Millions of Web developers around the globe have to pick the best-matching type and property from schema.org. Using a less widely understood yet more precise term may actually worsen the quality of markup, or reduce the amount of it, or both. Martin > On 11 Jun 2018, at 10:17, Simon.Cox@csiro.au wrote: > > My read of all this is that > > schema.org:Event subClassOf natural-language:Event . > > You need to recognize that the name of the concept in schema.org is more specialized that the use of the word 'Event' in all English-language usages. > > The upper-ontologies resolve this by using a more technical word - like occurrent or perdurant - which includes all time-bounded things including both planned events, unplanned incidents, things that move around, as well as other kinds of interactions. But even these attempts to define universals must be taken with a pinch of salt. Classification is contextual - the 'same thing' may be classified differently in different applications. The main application of schema.org is SEO, so the classification system is tuned to that. > > > Simon J D Cox > Research Scientist > Land and Water > CSIRO > E simon.cox@csiro.au T +61 3 9545 2365 M +61 403 302 672 > Physical: Reception Central, Bayview Avenue, Clayton, Vic 3168 > Deliveries: Gate 3, Normanby Road, Clayton, Vic 3168 > Postal: Private Bag 10, Clayton South, Vic 3169 > people.csiro.au/C/S/Simon-Cox > orcid.org/0000-0002-3884-3420 > researchgate.net/profile/Simon_Cox3 > From: Anthony Moretti <anthony.moretti@gmail.com> > Sent: Monday, 11 June 2018 8:26 AM > To: Dan Brickley > Cc: Peter F. Patel-Schneider; Richard Wallis; public-schemaorg@w3.org > Subject: Re: Schema.org and OWL > > Yep I definitely understand it doesn't imply disjointedness. But subclassing is answered by the "some or all" test, and I'd argue that all instances of these types are events. For example, is there an example of a trip that is not an event? > > And yep I completely agree, some not all books have current offers for sale/trade, so books are not a subtype of product. And I completely agree with the second example too, some not all things being offered are creative works. For example, an orange being offered for sale is a product but not a creative work, thus not all products are creative works. So I don't think those are examples of tradeoffs by schema.org, they're completely correct modeling by schema.org. > > If things like trips aren't considered events then strange behavior or additional work for people will result. For example a reasoner might look for all events to display on a calendar, but with the current structure all trips will have to be additionally typed as events, when in my view it should be implicit from already being classified as a trip. > > On Sun, Jun 10, 2018 at 2:38 PM Dan Brickley <danbri@google.com> wrote: > The lack of a "subclassof" relationship doesn't mean we consider the types disjoint. You might similarly argue that all books are products, or that all products are creative works. While we respect the efforts of the ontologies community, Schema org is a bit different in the tradeoffs we have chosen to make. The looseness may be frustrating but it has also helped us grow this thing incrementally. > > Dan > > > On Sun, 10 Jun 2018, 14:28 Anthony Moretti, <anthony.moretti@gmail.com> wrote: > If we believe schema.org will be used long into the future then even if changes are hard we should try to make them right? > > Arguing by word origins, even the word "action" comes from "event". Using the dictionary on Google: > > "action" > "act" > "actus" (Latin) - event, thing done > > Arguing by structure, Actions on schema.org duplicate the startDate and endDate properties (startTime and endTime). Additionally, Actions have a state property onschema.org but Events do not, yet all action states can all be derived from more general event states. So putting Action, or any type with a startDate or endDate, as a subtype of Event actually removes redundancy. > > A simple test to determine whether something is an event or not is can it be put on a calendar and look normal? I think instances of all the types I mentioned can: > > Event > Action > Course > Offer > ParcelDelivery > Relationship > Trip > > Anthony > > On Sun, Jun 10, 2018 at 11:38 AM Dan Brickley <danbri@google.com> wrote: > Yeah, we decided against cramming Actions under Event a long time ago, even if it has a certain appeal; I' d say same goes for the other event-ish types. > > On Sun, 10 Jun 2018, 11:24 Peter F. Patel-Schneider, <pfpschneider@gmail.com> wrote: > It's very difficult to determine what things are instance of schema.org Event > because the guidance is so slim, particularly as we are not supposed to take > into account most of the information on http://schema.org/Event. My view is > that even if we are supposed to ignore everything after "certain time and > location", then schema.org Event should be read narrowly, excluding things > that don't have a certain time or don't have a certain location. This would > rule out Action, ParcelDelivery (as the entire delivery process), and Trip, as > these generally take place over several locations. Offer is ruled out > because it often doesn't have a location at all and its temporal information > doesn't have to do with its "happening". Course is also ruled out because it > is more than a sequence of lectures. > > > One could take an expansive view of schema.org Event, perhaps saying that its > instances are anything that has a temporal component and a spatial > component. So Bill Murray is an instance of schema.org Event via his birth > date and his location on Earth, as is his membership in the cast of > Ghostbusters. But then what things aren't instances of Event (besides > platonic entities like 3, if you assume that there are any truly platonic > entities)? > > > Is there a middle ground? Formal ontologists have attempted to create > one---dividing the universe into endurants and perdurants. But I don't see > that schema.org Event is getting at the meaning of perdurant. > > > peter > > > > > On 06/10/2018 12:32 AM, Anthony Moretti wrote: > > Let's get the easy ones out of the way before looking at Roles. The > > following are obviously subtypes of event, I might create an issue on GitHub > > to get more feedback: > > > > Event > > > > Action > > > > Course > > > > Offer > > > > ParcelDelivery > > > > Trip > > > > > > I'll argue for Role now. In my view, it would be easier to understand if the > > terminology was Relation, or Relationship, rather than Role. > > > > An event is either: > > > > * A period of time (e.g. Jun 9, 2018, 12 PM - 1 PM) > > * The period of time that a statement, explicit or implicit, is true (e.g. > > My run this morning = Anthony isRunning True) > > > > To be clear, having the same statement exist with different periods of > > validity can still be consistent: > > > > * Anthony isRunning True (Jun 9, 2018, 12 PM - 1 PM) > > * Anthony isRunning True (Jun 8, 2018, 12 PM - 1 PM) > > > > If you look at temporal databases > > <https://en.wikipedia.org/wiki/Temporal_database> every fact has a valid > > time <https://en.wikipedia.org/wiki/Valid_time> - the time period during > > which the fact is true. Therefore if you reify any fact or relationship you > > produce something with start or end times, therefore producing an event. > > > > In schema.org <http://schema.org>: > > > > 1. All Roles are reified relationships. > > 2. Therefore all Roles have "valid times". > > 3. Therefore all Roles have start or end times. > > 4. Therefore all Roles are events. > > > > If the terminology was updated it would look like: > > > > Event > > > > Relationship, or Relation (currently Role) > > > > > > Regarding Bill Murray, what is being modeled if not the fact that Bill > > Murray was involved as an actor during its filming, and so for a period of time? > > > > And yep, I agree that you can define an entity and separately define an > > event that represents its period of existence. But you can't do that for an > > event, and in like fashion you can't do that for a statement either, it just > > wouldn't make sense for either right? > > > > Anthony > > > > > > On Sat, Jun 9, 2018 at 6:52 PM Peter F. Patel-Schneider > > <pfpschneider@gmail.com <mailto:pfpschneider@gmail.com>> wrote: > > > > Yes, many Roles in schema.org <http://schema.org> are events, and could > > be so modelled. But not > > all, in my opinion, or at least not usefully. > > > > > > For example the relationship between Bill Murray and Ghostbusters is not an > > event. It was, perhaps, initiated by a bunch of events, namely a > > sequence of > > acting performances that were captured and edited together to form the > > movie. > > However, that bunch of events is not the relationship being modelled. > > > > > > It's just like one might model people as their lives, i.e., an event that > > plays out (roughly) from conception to death. But I find it useful to > > distinguish between a person and their life event. > > > > > > peter > > > > > > > > On 06/06/2018 09:01 PM, Anthony Moretti wrote: > > > I agree that the first problem to do with strings can be solved by boxing > > > them into anonymous individuals. On the second problem to do with Roles I > > > agree that you can't do a simple "subjectOf" transform because Role is not > > > in the domain of "about", but maybe defining a new property such as > > > "participantIn" (domain Thing, range Role) and using it wherever a Role is > > > the object of a statement might keep about the same level of meaning (not > > > very good) but at least make it consistent OWL? So your example would be > > > transformed to: > > > > > > DJT : Person > > > > > > participantIn ? : Role > > > > > > spouse MM : Person > > > > > > startDate : 1993 > > > > > > endDate : 1999 > > > > > > participantIn ? : Role > > > > > > spouse IT : Person > > > > > > startDate : 1977 > > > > > > endDate : 1992 > > > > > > > > > A more general solution using guidance > > > from https://www.w3.org/TR/swbp-n-aryRelations/ and with better meaning > > > might be: > > > > > > DJT : Person > > > > > > isSpouseFor ? : MarriageRelation > > > > > > spouse DJT : Person > > > > > > spouse MM : Person > > > > > > startDate : 1993 > > > > > > endDate : 1999 > > > > > > isSpouseFor ? : MarriageRelation > > > > > > spouse DJT : Person > > > > > > spouse IT : Person > > > > > > startDate : 1977 > > > > > > endDate : 1992 > > > > > > > > > Your question about movie roles is also solved by this, for example: > > > > > > Bill Murray : Person > > > > > > isActorFor ? : PerformanceRelation > > > > > > actor Bill Murray : Person > > > > > > movie Ghostbusters : Movie > > > > > > characterName : Dr. Peter Venkman > > > > > > startDate : 1983 > > > > > > endDate : 1984 > > > > > > > > > Class hierarchy: > > > > > > Event > > > > > > Relation > > > > > > MarriageRelation > > > > > > PerformanceRelation > > > > > > > > > I think confusion might come from the many English meanings of "role": > > > > > > Type of role, e.g. Quarterback - no start or end dates. > > > Instance of a role, e.g. 49ers quarterback 1979-1992 - potential start > > > and end dates. > > > Character, e.g. Hamlet - no start or end dates. > > > > > > > > > So I'm still lead to believe that Role/Relation and several other types in > > > schema.org <http://schema.org> <http://schema.org> are subtypes of > > Event because their instances > > > always have potential start and end dates: > > > > > > Event > > > > > > Action > > > > > > Course > > > > > > Role/Relation > > > > > > Offer > > > > > > ParcelDelivery > > > > > > Trip > > > > > > > > > Anybody else think this too? I'm a bit new to this so hope I'm not putting > > > something too strange out there! > > > > > > Anthony > > > > > > > > > On Wed, Jun 6, 2018 at 3:20 PM Peter F. Patel-Schneider > > > <pfpschneider@gmail.com <mailto:pfpschneider@gmail.com> > > <mailto:pfpschneider@gmail.com <mailto:pfpschneider@gmail.com>>> wrote: > > > > > > The short-circuiting of Role in schema.org <http://schema.org> > > <http://schema.org> is > > > certainly outside of > > > schema.org <http://schema.org> <http://schema.org>. OWL has > > nothing that can do the right > > > thing with the following > > > schema.org <http://schema.org> <http://schema.org> data > > > > > > DJT : Person > > > spouse ? : Role > > > spouse MM : Person > > > startDate : 1991 > > > endDate : 1999 > > > spouse ? : Role > > > spouse IT : Person > > > startDate : 1977 > > > endDate: 1992 > > > spouse MT : Person > > > particularly in conjunction with > > > spouse > > > domain Person > > > range Person > > > > > > Except that you could turn *all* property values (except maybe for > > those > > > properties that are guaranteed never to have a Role intermediary) into > > > events > > > *before* they got to OWL. This would look something like: > > > > > > DJT : Person > > > spouseEvent ? : SpouseEvent > > > spouse MM : Person > > > startDate : 1991 > > > endDate : 1999 > > > spouseEvent ? : SpouseEvent > > > spouse IT : Person > > > startDate : 1997 > > > endDate : 1992 > > > spouseEvent ? : SpouseEvent > > > spouse MT : Person > > > spouse range Person > > > domain SpouseEvent > > > spouseEvent range SpouseEvent > > > domain Person > > > SpouseEvent sub Event > > > startDate range Date > > > endDate range Date > > > > > > Hah! That turned out better than I thought it would, but I view the > > > need for > > > the transform as showing that schema.org <http://schema.org> > > <http://schema.org> Role is > > > outside the purview of OWL > > > because it requires major surgery to the underlying ontology (as > > opposed > > > to a > > > transform to handle "strings as things" that just replaces such > > strings > > > with a > > > blank node connected to the string by some sort of description > > property > > > which > > > doesn't require such surgery). > > > > > > And then what do you do about movie roles? They shouldn't have start > > > and end > > > dates, or even have the possibility of start and end dates. > > > > > > peter > > > > > > > > > > > > On 06/06/2018 02:50 PM, Anthony Moretti wrote: > > > > Peter, when you say "roles fall outside the purview of OWL" is > > it possible > > > > that Role should be a subtype of Event, thus requiring no special > > > treatment? > > > > I say that because individual roles have potential start and end > > dates. > > > > > > > > If you continue that thinking you end up with several types that > > might be > > > > considered subtypes of Event: > > > > > > > > Event > > > > > > > > Action > > > > Course > > > > Role > > > > Offer > > > > Parcel delivery > > > > Trip > > > > > > > > > > > > Anthony > > > > > > > > On Wed, Jun 6, 2018 at 7:42 AM Peter F. Patel-Schneider > > > > <pfpschneider@gmail.com <mailto:pfpschneider@gmail.com> > > <mailto:pfpschneider@gmail.com <mailto:pfpschneider@gmail.com>> > > > <mailto:pfpschneider@gmail.com <mailto:pfpschneider@gmail.com> > > <mailto:pfpschneider@gmail.com <mailto:pfpschneider@gmail.com>>>> wrote: > > > > > > > > OK, it appears that you are trying to produce an OWL > > ontology that > > > would > > > > accept schema.org <http://schema.org> <http://schema.org> > > <http://schema.org> > > > documents that would accept > > > > unexceptional schema.org <http://schema.org> > > <http://schema.org> <http://schema.org> > > > > documents (under some loose definition of what makes an > > unexceptional > > > > schema.org <http://schema.org> <http://schema.org> > > <http://schema.org> document). > > > > > > > > > > > > But you are running up against problems with "strings as > > things" and > > > > Roles. > > > > > > > > > > > > Look at about and subjectOf. Any use of a string as a > > value for > > > about is > > > > going to make that string an instance of two OWL Classes > > (the range of > > > > about, > > > > which includes Text, and the domain of subjectOf, which > > doesn't). > > > This > > > > is not > > > > permissible in OWL. > > > > > > > > > > > > Using a role as a value for subjectOf will make the role node an > > > instance of > > > > the domain of about, which doesn't include Role. Roles fall > > > outside the > > > > purview of OWL. > > > > > > > > > > > > What I think has to be done for "strings as things" is to > > > preprocess them as > > > > an anonymous node with the string as the value of some > > description > > > > property. > > > > Roles need to be short-circuited and the role values > > eliminated. > > > If the > > > > role > > > > values are to be retained then some sort of fancy > > reification has > > > to be > > > > introduced. > > > > > > > > > > > > But a lot of this is guesswork, as it is unclear just what > > "strings as > > > > things" > > > > and Roles mean in schema.org <http://schema.org> > > <http://schema.org> > > > <http://schema.org>. (I haven't looked > > > > closely at the use of URLs > > > > as stand-ins for objects but my guess is that that is similar to > > > "strings as > > > > things".) > > > > > > > > > > > > > > > > OWL makes a distinction between object and data values, and this > > > distinction > > > > has to be carried through to the ontology, where there are > > classes and > > > > datatypes. As far as I can tell Text should be a datatype. > > > > > > > > > > > > > > > > > > > > peter > > > > > > > > > > > > > > > > On 06/06/2018 07:09 AM, Richard Wallis wrote: > > > > > Just the fresh pair of experienced eyes I needed! - Thanks > > @Thomas > > > > > > > > > > @Danbri - not “wasting a lot of time agonising", in simple > > terms > > > I’m just > > > > > [selfishly] trying to get a useful version to load > > into Protégé and > > > > > hopefully helping a few others at the same time. > > > > > > > > > > @Peter thanks for your thoughtful questions. As you can see > > > from the > > > > above > > > > > the prime simple objective is to get it visible > > in Protégé. Beyond > > > > that, it > > > > > is to capture the class and property hierarchy of Schema.org > > > including > > > > > the multiple domain/ranges of properties defined [In > > Schema.org] > > > using > > > > > domainIncludes & rangeIncludes, in a way that a tool > > > like Protégé can cope > > > > > with. As for ranges, it includes the assumption that > > properties, in > > > > > addition to the defined range(s), also have Text, URL, and > > Role > > > > included in > > > > > their range. > > > > > > > > > > Once I’ve done it, I want to add the simple generating code to > > > the Schema > > > > > scripts run at release time so that it can be kept up to date. > > > > > > > > > > > > > > > ~Richard. > > > > > > > > > > Richard Wallis > > > > > Founder, Data Liberate > > > > > http://dataliberate.com > > > > > Linkedin: http://www.linkedin.com/in/richardwallis > > > > > Twitter: @rjw > > > > > > > > > > On 6 June 2018 at 14:45, Peter F. Patel-Schneider > > > > <pfpschneider@gmail.com <mailto:pfpschneider@gmail.com> > > <mailto:pfpschneider@gmail.com <mailto:pfpschneider@gmail.com>> > > > <mailto:pfpschneider@gmail.com <mailto:pfpschneider@gmail.com> > > <mailto:pfpschneider@gmail.com <mailto:pfpschneider@gmail.com>>> > > > > > <mailto:pfpschneider@gmail.com > > <mailto:pfpschneider@gmail.com> <mailto:pfpschneider@gmail.com > > <mailto:pfpschneider@gmail.com>> > > > <mailto:pfpschneider@gmail.com <mailto:pfpschneider@gmail.com> > > <mailto:pfpschneider@gmail.com <mailto:pfpschneider@gmail.com>>>>> wrote: > > > > > > > > > > It's hard to say much about the file without knowing > > what it is > > > > supposed to > > > > > capture. > > > > > > > > > > > > > > > It is supposed to capture the class and property > > hierarchy and > > > > property > > > > > restrictions or schema.org <http://schema.org> > > <http://schema.org> > > > <http://schema.org> > > > > <http://schema.org>, but not necessarily in a > > > > > form compatible with RDFS > > > > > or OWL? > > > > > > > > > > > > > > > Is it supposed to faithfully encode the model theory of > > > schema.org <http://schema.org> <http://schema.org> > > > > <http://schema.org> > > > > > <http://schema.org> in OWL? > > > > > If so, where is the document for this theory? > > > > > > > > > > > > > > > Is it supposed to capture "strings as things" or Roles? > > > > > > > > > > > > > > > How does it view property domains and ranges? As > > axioms? > > > As strict > > > > > constraints? As soft constraints? > > > > > > > > > > > > > > > I would also move from rdf/xml to turtle, which is > > easier to > > > write and > > > > > easier > > > > > to read. > > > > > > > > > > > > > > > peter > > > > > > > > > > > > > > > > > > > > On 06/06/2018 05:05 AM, Richard Wallis wrote: > > > > > > Calling folks with more OWL experience than me! > > > > > > > > > > > > The schema.org <http://schema.org> > > <http://schema.org> <http://schema.org> > > > <http://schema.org> > > > > <http://schema.org> site has an OWL > > > > > definition file that has > > > > > > not been maintained since April > > > > 2014: http://schema.org/docs/schemaorg.owl > > > > <http://schema.org/docs/schemaorg.owl>. > > > > > > Also the structure and syntax of the file needs some > > > attention. > > > > > > > > > > > > To help with the occasional questions about accessing > > > processable > > > > > > representations of the vocabulary; to attempt to > > close an > > > issue > > > > (#1611 > > > > > > <https://github.com/schemaorg/schemaorg/issues/1611 > > > > > > > <https://github.com/schemaorg/schemaorg/issues/1611>>); and to > > > > help with a > > > > > > personal project, I have had a look at producing an > > up to > > > date, > > > > improved, > > > > > > maintainable version of the file. > > > > > > > > > > > > My first attempt can be downloaded/viewed > > > > > > > > > here: https://s3.amazonaws.com/rjwPublicData/public/schemaorg.owl > > > > > > > <https://s3.amazonaws.com/rjwPublicData/public/schemaorg.owl> > > > > > > > > > > > > I am looking for comments, suggestions, and help > > around a few > > > > aspects of > > > > > > this work in progress: > > > > > > > > > > > > * Is it generally ‘a good owl file’ > > > > > > > > > > > > * Should it contain more/less info about > > the vocabulary and > > > > its terms > > > > > > > > > > > > * Specifically with reference to domainIncludes and > > > rangeInclude - > > > > > mapped > > > > > > to rdfs:domain & rdfs:range with > > owl:unionOfcollections: > > > > > > > > > > > > o Is this the best/only way to represent multiple > > > domain & > > > > > ranges for > > > > > > an objectproperty? > > > > > > > > > > > > o Have I got the syntax correct? > > > > > > > > > > > > * Several people use Protégé > > > <https://protege.stanford.edu/> as a > > > > > tool for > > > > > > this kind of effort - I am trying to identify > > what syntax, > > > > will enable > > > > > > this tool to recognise the multiple > > domain/ranges when > > > importing > > > > > this file. > > > > > > > > > > > > If anyone out there with more OWL experience than me > > (not > > > > difficult), > > > > > could > > > > > > spend a few minutes taking a look at this and > > commenting, it > > > > would be > > > > > > greatly appreciated. > > > > > > > > > > > > ~Richard > > > > > > > > > > > > Richard Wallis > > > > > > Founder, Data Liberate > > > > > > http://dataliberate.com > > > > > > Linkedin: http://www.linkedin.com/in/richardwallis > > > > > <http://www.linkedin.com/in/richardwallis> > > > > > > Twitter: @rjw > > > > > > > > > > > > > > > > > > > > > > > > > > > > > >
Received on Tuesday, 12 June 2018 08:07:07 UTC