event should have ptr and idno in its content model

[jamescummings]

I'm currently at a project meeting where we are talking about events as first-order entities, akin to but separate from named entities such as people/places/orgs/works. Currently <event> allows <link> and <linkGrp> but it would be extremely useful to have a way to point from these events to other entities or external taxonomies with a way to classify each of those pointers. This would need <ptr> with a @type attribute or similar. (Rather than just using @corresp or something.) <link> is intended for linking two IDs together, so not suitable for this. People often work around this by having a <bibl> with only a <ptr> inside, but this is silly since it isn't necessary a bibliographic entry.

Proposal: add either <ptr> or model.ptrLike to the content model of <event>.

[martindholmes]

Yes please!

[jamescummings]

Based on viewing Event as a first-order named entity in parallel to person/place/org I examined the content models of those three for comparison. They all have idno, link, and linkGrp, but not ptr. Event has link and linkGrp but not idno and ptr.

Any event, as person/place/org may well have an identification number of various sorts. All named entities may have a need to point to other sources of information, definitions of that entity, etc. and these should not be done merely with an attribute (like corresp) because the nature of the pointing can rightly be classified (i.e. it needs att.typed). This is precisely what ptr is for.

Since a full rationalisation of named entities to have broadly comparable content models is something council has already decided is a good idea, but undertaking a full revision is a monumental task that we've demurred on for some time, I'd like to try a more incremental approach. My principle being that any proposed smaller changes that bring the content models of named entities into greater cohesion of content models and doesn't preclude later rationalisation are probably a good ideas. We shouldn't need to wait for the perfect solution to implement smaller ones..

So the logical possibilities to choose from then are:

a) Just add ptr to event's content model b) Just add idno to event's content model c) Add ptr and idno to event's content model d) Add ptr to event/person/place/org e) Do something different (please comment) f) Add idno to event's content model and ptr to event/person/place/org e) Do none of these, keep the status quo.

I'm willing to undertake the necessary changes if approved. I would vote for 'f' since it provides the most parity and answers numerous use cases I've heard for awhile. I'll ask council (and others who wish) to vote/comment here. (A thumbs up will be taken as agreeing with proposal 'f'.)

[raffazizzi]

I agree overall, but currently there is a clear hierarchy between person/place/org and event. This is already clear from event's desc: "contains data relating to any kind of significant event associated with a person, place, or organization". Why was it decided that en event is not a first-order entity? This is importantly reflected by the fact that event can be contained by person/place/org, but these three cannot contain each other.

So, I think it makes sense to add ptr to event's content model (option a), but I'm unsure about idno. It seems tidier to only allow event/idno if it's an event defined within an eventList rather than within a person/place/org. If not, events will be the only entities that can be defined (by unique identification) within other entities.

[laurentromary]

That would make things too complicated. Whether an event, say the Siege of Maubeuge, is referred to as part of a place (Maubeuge), a person (Governor Joseph Fournier) or a list of WWI events, you will have the same reason to point to its identifier in Wikipedia for instance (or any other database). Preventing idno to be used in some of those usages would not be optimal to cover a variety of possible use cases.

[raffazizzi]

@laurentromary I would certainly want to have event be able to point to outside resources and that made me think that ptr was the right tool for the job (and often is). I forgot that idno can be used for a similar purpose, e.g. by including a database identifier. Thanks for clarifying this for me!

So I change my vote to (f).

[martindholmes]

I like f) too, although I wish there were a simple class-based solution. I think the definition of <event> as cited by @raffazizzi above should be changed; there's no reason why events shouldn't be first-order entities. That's surely what we're heading towards.

[lb42]

Can anyone suggest a plausible example of an event which is NOT associated with some person place or org? If not, the case for conferring first class status on it is clearly weakened, if not entirely adumbrated...

[duncdrum]

FWIW, and a change of our usual dynamics, I m with Lou on this one. I have no problems with the concept of standalone place, name, or date entities. But what would be an example of an event entity that does not depend on at least one of those?

[laurentromary]

Read my example: in historical descriptions, you can describe sequences of events which (I would say conversely ;-)) subordinate a variety of places, persons, materials, what have you. Having these events under the specific angle of a person is accidental. e.g. Governor Joseph Fournier is part of the Siege of Maubeuge and you would not force describing the event as part of the (quite boring) life of Governor Joseph Fournier. In the same way, even if we speak of the Siege of Maubeuge, the event is not necessarily to be seen as part of the specific description of the (still very nice) city of Maubeuge, right after a description of the béguinage des Cantuaines.

[martindholmes]

The plot of a novel may be a sequence of events. If you're not interested in the characters or locations, you just want an event list, which perhaps you would categorize according to some taxonomy for analysis.

[laurentromary]

The description of a chemical process in a scholarly paper (yes, there are some encoded in TEI...)

[duncdrum]

idgi how is a sequence of fictional or factual events not captured by @where and att.datable on the event elements? Whats the point of encoding the 'Siege of Maubeuge' or the battle of mordor if there is no <place> for Maubeuge or mordor . Your problem with sequences of events, seems more related to the placement of labeled listEvents, then with the first class entity status of event, no? Sounds like c) would do the job.

@laurentromary the description in a paper implies bibl to me. For the process itself to be an event it would need to happen and be observed at a certain time and place, s.a.. If we talk about generic types of chemical processes like photosynthesis taxonomy seems more suitable.

[bansp]

@lb42 This sounds like you're saying that the TEI should restrict the possible kinds of area- and perspective-specific ontologies expressible in it rather than allow to (shallowly) encode most forms of such ontologies that one might care to encode in the TEI. (This is said on the assumption that I think we share, that the TEI is suitable or at least usable as a language of shallow ontology description.)

Or are you making an ontological statement that events can never be taken as first-class entities, even in (let's say, just for the sake of argument) "naive" conceptualizations?

[martindholmes]

@duncdrum Let's imagine I'm encoding a set of novels, for the purpose of analyzing the types of events that occur in the novel. There are no dates -- it's a novel -- and I don't care about characters. I care about identifying a point in the text where something occurs, linking that to an event description (marriage, funeral, outbreak of war), and then I link these events to categories in a taxonomy of event types. No people, no places, no dates -- just events, which are perfectly good first-order entities.

[jamescummings]

@lb42 @duncdrum I agree with Martin's use cases and @laurentromary's explanation and the kind of one which I was thinking of was where a source has a list that is merely a list of events. Yes the Siege of Maubeuge could be said to be about that place but where your list is an eclectic mix of things. Lets say you have a source which you want to describe with listEvent where each entry includes an id number as well as many references to further info with a mixture of different places, times, people, or organisations involved. It is silly to expect people to embed all of these in separate entities.

Event001: Siege of Maubeuge, more details here Event002: Convention of the constitution of the USA (happens over range of dates and technically not in the same place) more details at this URL. Event003: Rowe vs Wade, referred to as an event, took place over time in a single place, but is probably the date of judge's ruling. Event004: Great North Exhibition. Took place in multiple locations Event005: The London 2012 Olympics (didn't take place all in London) Event006: WW2 happened lots of places and not just to specific people or orgs

I don't deny that events are things in the real or fictional world that happen to combinations of people/places/orgs but people also refer to them entirely separately from those people, places or orgs. And there are historical gazetteers and authority files listing those events that it would be useful to point to.

I've encountered several projects that want to treat events as first order objects in themselves and shoehorn ptrs and idno into a bibl when it really isn't appropriate.

[emylonas]

I like the idea of event as a first order entity, for many of the reasons above. We shouldn't mandate how a project might prioritize the kinds of things it identifies - if people are central, then events might be subordinate to them, but conversely, if events are more significant (a chronicle?), then even though people and places are mentioned, they are constituent parts.

So making their structures parallel seems reasonable. It would be great to change event to a first order object, but that is a different issue.

Vote for f.

[ebeshero]

I agree with James, Raff, Martin, and Elli--the use cases and capacity for a project to be flexible are paramount, so yes, I support f. This comes at a moment, by the way, when I'm reading ecocriticism with my literary theory students, in which human concerns are "embedded" in landscapes and shown to be contingent on nonhuman conditions. An ecocritical markup method might well make humans contingent on events like natural disasters.

Ecocritically, Elisa

On Wed, Mar 28, 2018 at 11:33 PM, Elli Mylonas notifications@github.com wrote:

I like the idea of event as a first order entity, for many of the reasons above. We shouldn't mandate how a project might prioritize the kinds of things it identifies - if people are central, then events might be subordinate to them, but conversely, if events are more significant (a chronicle?), then even though people and places are mentioned, they are constituent parts.

So making their structures parallel seems reasonable. It would be great to change event to a first order object, but that is a different issue.

Vote for f.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/TEIC/TEI/issues/1756#issuecomment-377110424, or mute the thread https://github.com/notifications/unsubscribe-auth/AD1BtiefN72D_gyncmjSz55JL2-_7fLvks5tjFYUgaJpZM4S3cQ7 .

-- Elisa Beshero-Bondar, PhD Director, Center for the Digital Text | Associate Professor of English University of Pittsburgh at Greensburg | Humanities Division 150 Finoli Drive Greensburg, PA 15601 USA E-mail: ebb8@pitt.edu ebb8@pitt.edu Development site: http://newtfire.org

[sydb]

I see no reason why <event> should not be considered a first-order object, and suspect that if we asked an historian she would be miffed that they aren’t already. That said, the thing I would like to avoid is providing multiple methods of encoding the exact same thing inside the data structure. Why allow both <idno> and <ptr> for the exact same purpose? (And someday someone is going to suggest @ref on <person>, <event>, <place>, <org>, and <nym>.) My instinct is we should pick one of them, and even have a suggested value of @type that means “go here for more information about the very thing represented by this TEI object”, and another that means “go here for further interesting reading that you would likely find interesting”.

And, as the bloke who is supposed to be rationalizing the person/place/org/event content models (as soon as I’m done rubbing), I am all in favor of James’ “incremental approach” idea, but warn that it could be quite difficult to do.

[jamescummings]

@sydb I see idno ptr and ref attribute as all separate things. An event or other named entity may have one or more identification strings. It may also have things one wants to point to as sources of more information or description that can be classified in different ways. It may also have an equivalent that you might point to with a ref attribute.

[jamescummings]

Labelling as Status: Needs Discussion for now since there were a number of challenges though a number of commentors seem broadly in favour.

My view of an event that is separate from an individual person/place/org is something like "World War Two". Suppose we have a list of events we want to encode in a <listEvent> and that is one of them. It doesn't correspond to any particular person/place/org (thus the desc of <event> should be loosened). In the local encoding project's system it may have an identifying number they wish to record. Or they may wish to use non-URI identification numbers associated with the event. (I don't know, maybe something like its dewey decimal code? I'm sure there are better examples.) It may also wish to point to either equivalences or further information in sources which could be classified (e.g. using @type on <ptr>) by the nature of the pointing/resource.

[duncdrum]

@jamescummings My understanding is that we are debating at least three issues:

• About ptr/idno/@ref I stick with c).

The other two concern:

• the ontological status of events and where in the tei hierarchy listEvent can occur;
• the question of events occurring in multiple or no place (and time), seems again a separate issue with the content model for @where and maybe @type;

I'd suggest to create separate tickets for the final two so the first doesn't get bugged down by me doing the Ophelia to what is dreamt of in Piotr's philosophy 💃

[raffazizzi]

F2F subgroup: It seems like the majority of posters here agrees with proposal f) and we think it should be done.

[martinascholger]

Council (F2F Washington) agrees with proposal f).

[martinascholger]

Tested with lastSuccessfulBuild, works as proposed in f). Can we close the ticket, @jamescummings?

[jamescummings]

This is done, sorry, was only waiting for build to be finished on jenkins but it was broken at the time. All good.