[Data-modeling] "Mentions" and "Mentioned In"
Scott Meyer
sm at metaweb.com
Sat Feb 7 00:54:47 UTC 2009
Ed Laurent wrote:
> For these other examples, it seems you would want to keep the context as
> well as document the mention.
>
> For general mentions, your "Mentions" and "Mentioned In" as reciprocal
> properties sounds good. I'd suggest including "Written work" in your
> "Work mentioned" type (although in this case I would call it "Mentioning
> work" or "Source of mention" or...). Your "Topic mentioned" type could
> be included in various other types that defined their context. For
> example, "Location mentioned in fiction" could include "Topic mentioned"
> and "Location" with a link to "Mentioning work of fiction" (or something
> similar), that includes "Mentioning work" and "Book". The literary genre
> of "Fiction" would need to be added by hand or script afterwords though.
>
> Just some thoughts. In any case, my opinion is that these are distinct
> types that are worth modeling but should not be included as properties
> of "Written work" (which is included in "Book") because that would be
> too limiting for them. You want to include them in your more
> use-specific types so they need to stand alone.
>
> -Ed
>
> P.S. There are several "mentions" in the Aubrey-Maturin base:
> http://aubreymaturin.freebase.com/ but they don't include a general
> "Mention" type. They probably should based on this discussion.
Ah, you've caught us in mid schema-design.
The type specific mention types (ie. character mention, ship mention, etc.)
were built that way because they looked good in the somewhat limited
table view that was current when that schema was designed. Given that
those constraints have been removed, I think that it makes much more
sense to allow applications to constrain the type of the thing
mentioned (if they so desire). Also, it is much better to define
properties such as quote and circumstances once (and have "sub" classes
delegate to them) than to create N different properties with identical
meanings.
I'm sitting on a more generic mention type doesn't constrain the type
of the thing mentioned. It has properties for:
what (/common/topic)
The thing being mentioned
circumstances
How "what" was mentioned
quote
A quote from the text
chapter
A chapter object, could also be text
page
{cvt of: edition, page number}
The last four fields specify (with some redundancy) a location in a
text.
So how about creating a generic mention type which would be a superset
of all types of mention. The thing mentioned (what) is universal, but
the location of the mention varies from domain to domain
(chapter-section-paragraph-line, book-verse, act-scene, time from start)
Define as many as we can think of and let users fill in whichever are
appropriate to their domain.
In the specific case of aubreymaturin, we'd create our own mention
type delegating what, circumstances, and quote and adding our own
chapter and page structure since those are specific to the base
which explores a small set of books in greater-than-usual detail.
-Scott
More information about the Data-modeling
mailing list