Reification, the ability to treat a relation as a resource, is very well understood (the difference between an N order logic and an N+1 order logic is covered in most first year logic courses) but not yet well implemented in RDF. Arbitrary levels of reification (as implemented in Wolfram) are unlikely to be directly useful in many day-to-day library settings.

None-the-less defining library metadata in terms of piecewise-defined directed graphs is important for ease of understanding and broad consistent application. After all, if the first-order data is not of sufficient quality, all the logic in the world isn't going to fix that.

cheers

stuart

--
I have a new phone number: 04 463 5692

Sent: Tuesday, 28 April 2015 7:52 a.m.
Subject: Re: [BIBFRAME] Linked data and directionality

I don’t think that the limited number of graph concepts adopted by RDA designers can truly support what seems to be the “mixed graph" structure described by Stuart.

Given this basic definition:  http://en.wikipedia.org/wiki/Mixed_graph, and seeing that a heavyweight graph manipulation tool like Mathematica: http://reference.wolfram.com/language/ref/MixedGraphQ.html only just implemented mechanisms to test for and manipulate mixed graphs,

Setting policies that piecewise define directed graphs only will minimize graph processing hassles (we are not doing scheduling problems). This would require for example, that we set policies that resolve what are asserted to be bidirectional relationships explicitly into their component directed parts (I’m looking at you, RT).

Not A Bad Idea, That –

The motivation for mixing graph structures is well understood –  viz “expressive power":

But!

A more intensive commitment to graph structuring and subsequent operations on the structure is implied – one that is not supported by moments in this thread.

Prior Art – Wolfram Mathematica represents graphs in a much more complex fashion than what we see in RDF: each vertex and edge / node and link can be assigned properties, labels, UniqueIDs and even other graphs. This allows graphs to be easily manipulated by functions that handle directness and nondirectedness completely behind the scenes.

RDA, RDF, and BIBFRAME do not possess such graph processing machinery, hinting at likely inconsistent computational treatments at scale.

Ron Murray

Date: Monday, April 20, 2015 at 3:52 PM
Subject: Re: [BIBFRAME] Linked data and directionality

There are several points here:

a) SPARQL and other query languages let you search / query relationships in either direction.

b) Many vocabularies will have predefined OWL rules so that permit creatorOf can be automatically inferred from createdBy. (i.e. the reverse relationship has an explicit name)

c) Many vocabularies (implicitly or explicitly) have a preferred direction for the relationship (they might only define "creatorOf" or they might recommend implementers to use "creatorOf" and allow "createdBy" to be inferred when needed). This potentially reduces redundancy and increases clarity in data interchange.

d) Dates / Years are 'special' due to the complexity of different granularities of specification (sometimes it's a year of birth, sometimes it's an exact date, sometimes it's a time and date, sometimes the value is not in the Gregirian calendar, ...), but broadly similar functionality is available.

cheers

stuart

--
I have a new phone number: 04 463 5692

Sent: Tuesday, 21 April 2015 7:14 a.m.
Subject: Re: [BIBFRAME] Linked data and directionality

It does actually start with Shakespeare as a FOAF agent, and his age is a FOAF property. No?

-Bruce

Bruce J. Gordon
Audio Engineer
Audio Preservation Services - a shared service of the Harvard Library
Harvard University
Cambridge, Massachusetts 02138
U.S.A
tel. +1(617) 495-1241
fax +1(617) 496-4636

Isn't that direction an inference?
-Bruce

Bruce J. Gordon
Audio Engineer
Audio Preservation Services - a shared service of the Harvard Library
Harvard University
Cambridge, Massachusetts 02138
U.S.A
tel. +1(617) 495-1241
fax +1(617) 496-4636