[Core-Concepts] Entity Relationships

Tony Proctor tony at proctor.net
Thu Oct 9 05:16:01 CDT 2014


I've been trying to re-factor some of my existing [STEMMA] linkages in order to resolve some known long-standing problems. That's not going as nicely as I'd hoped but I wanted to jot down some of my notes in case they help others to visualise the problems at hand. The notes below are directly from a scrap of paper in front of me, and were an attempt to reduce the basic relationships to the most compact description I could think of.

I'm not suggesting this is the only way but it does encapsulate many of the nuances. Where I've mentioned "persona", I take it for granted that there is a direct equivalent for Place, Group, and whatever other subject entities are defined, and that including them explicitly would only obscure the representation.

As normal in software circles, I've used asterisk to indicate 0-to-many, and square brackets for an optional item.

    Source -> Eventa (*)

    Eventa -> Eventa-Property (*)

    Eventa -> Persona (*)

What this says is that any given source can identify 0-to-many eventa, i.e. references to some information with an associated date or date-range. Each eventa may yield a number of properties (aka attributes or "facts") about when and where this occurred. Each eventa may also yield a number of persona for person references (or equivalent for place/group references, etc).

    Eventa (*) -> [Event]

This indicates that several separate eventa, derived from different sources, may be references to the same conclusion-Event, i.e. the description that records the best where-and-when information, and which might be connected to other related Events in a hierarchical fashion.

    Persona -> Persona-Property (*)

    Persona (*) -> [Person]

This indicates that each persona may identify a number of properties concerning the associated reference (e.g. a name), and that multiple persona may be references to the same conclusion-Person  This persona-property relation probably should be 1-to-many, rather than 0-to-many, but I don't want to complicate this too much.

    Property -> [Value]

    Property -> [Persona]

This last bit indicates that each property has an optional value (such as a name, or an age), and an optional link to another persona (or place/group reference) - these being inclusive of each other. This implicitly means that a relationship to another subject reference can be modelled within the context of that particular time-and-place.


OK, if you're still with me [I know these pseudo-code examples make most people glaze over] then let's see how this pans out for a single hypothetical sentence found in some source.

    "In 1963, John Smith, of 10 Front St, brother of Simon Smith of Woodstown, married Ann Jones"

The sentence is contrived to be quite dense. It might be tempting for many genealogists to simply digest the information in their heads and record the individual parts on their conclusion entities, but then they'd lose the ability to back-track. Let's assume, though, that we're going to distinguish the raw references in the information from any association with our conclusion entities. Apologies for using "placea" but we don't have a term for place reference in this layer.

eventa E1
    property when=1963
end

persona P1
    property name=John Smith
    property residence=W1
    property relationship=P2.Brother
end

persona P2
    property name=Simon Smith
    property residence=W2
end

placea W1
    property name=10 Front St
end

placea W2
    property name=Woodstown
end

...etc...

Each of these evidential references can be independently associated with some equivalent conclusion entity. For instance, the eventa-E1 may be connected to a marriage Event, persona-P1 may be connected to a Person entity representing everything about a John Smith ancestor, and placea-W1 may be associated with a Place entity describing that house, it's location, pictures of it, etc. Hence, if Woodstown could not be identified reliably then the relationship could still be represented in this evidential layer.

STEMMA problems that I've been wrestling with in this area were that its equivalent of a persona were always associated with a conclusion-Person, and so couldn't be left dangling for unidentified or incidental references. Also, that a specific Property could reference a conclusion entity but not another persona/placea/etc. Finally, my relationship terms, such as Brother, could only be relative to an Event-specific Role (e.g. Head.Brother) rather than relative to another evidential reference.

(If nothing else then this post illustrates that we're desperately in need of a consistent set of terms for certain concepts)

    Tony
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://fhiso.org/pipermail/core-concepts_fhiso.org/attachments/20141009/3f4677ae/attachment.html>


More information about the Core-Concepts mailing list