Hi Phil,
> I've just posted a short update about the Learning Materials
> Application Profile scoping study at
> http://blogs.cetis.ac.uk/philb/2008/01/29/lmap-update/ which
> might interest some of you since it is related to CETIS
> metadata and will hopefully feed into the DC Education
> community's work.
>
> I'm particularly interested in comments on a straw-man domain
> model for learning materials in a repository, which is
> available at http://www.icbl.hw.ac.uk/lmap/domainModel.html
>
> Thanks in advance for all comments & discussion - Phil.
Slightly belatedly, an expanded version of what I wrote briefly on our
weblog last week [1].
First, I should say thanks for this. It looks like an interesting
starting point.
Only very briefly re the "Model for actions", I must admit I do wonder
in passing how well the OAIS Model really reflects all the functions of
a repository in the context of the current Web. After all, the OAIS
Model has been around since the late 90s, and I think our "repository"
Web apps have changed considerably since then. I'm thinking particularly
of the sort of "user-generated content" and "social" functions that are
being more widely adopted (consumer-as-provider/"prosumer", ratings,
reviews, ratings, tagging, "favourites" etc, formation of groups,
"friends" etc). But maybe the argument is that those are functions of
another system which sits alongside the repository rather than core
functions of the repository itself.
The remaining comments are on the "Model for objects".
(1) I think the idea of examining the use of FRBR is a very worthwhile
exercise. Whether it'll turn out to be the right tool for the job, I
don't know. I think the functional requirements need to be clearer
before we can decide that one way or the other. But I agree with you
that at least some learning materials are also scholarly works and the
use of FRBR in the former case should facilitate the task of
applications working with both the LMAP and the existing Scholarly Works
AP.
(2) Agent
I guess I'm a bit concerned that the current model uses a single Agent
entity type/class for what I think may really be better modelled as two
or more different types of thing.
FRBR itself doesn't have an Agent entity type/class; in SWAP the Agent
entity type/class was introduced as the super-class of the FRBR Person
and Corporate Body classes. The defining feature of these classes in
FRBR is, I think, that instances of these classes are (or were) capable
of action. The description of "Corporate Body" in FRBR section 3.2.6
says explicitly "an organization or group of individuals and/or
organizations acting as a unit."
So in LMAP, I can see that this fits the bill for the object of the
is-Created-By, is-Edited-By, is-Owned-By, and is-Published-By
relationships. In each of those cases the object could be the entity
called "Phil Barker" or the entity called "The JISC Repositories
Research Team".
But for the case of the is-Intended-For relationship, it seems to me
we're talking about a different entity type. It seems unlikely that in
these cases the object would be the entity called "Phil Barker" or the
entity called "The JISC Repositories Research Team". Rather the object
is more likely to be some class of agents e.g. your example of the class
of students in HE. This is not a FRBR Corporate Body; it doesn't act as
a unit.
For the case of the is-Used-By relationship, I'm not sure what the
intent is e.g. whether the object is indeed an individual or
organisation, or whether it is a class of agents. (Given that the
subject is a Copy, I think it's the former, but I'm not sure)
I think the very fact that there is (it seems to me) no overlap between
the entities which are objects of is-Created-By, is-Edited-By,
is-Owned-By, and is-Published-By relationships (e.g. "Phil Barker" or
"The JISC Repositories Research Team") on the one hand and the entities
which are objects of the is-Intended-For relationship ("students in HE",
"under-fives" etc) on the other suggests that we are really dealing with
two distinct entity types/classes. But probably the key point is that I
think thes instances will be described with different sets of attributes
and participate in different relationships. Sure, I'm jumping ahead of
myself but you can imagine a requirement to capture postal address or
email address for "Phil Barker" or "The JISC Repositories Research Team"
but that attribute would make no sense for "students in HE" or
"under-fives". And conversely you might want to say "first-year
undergraduates" are a subclass of "students in HE", which wouldn't
really apply to the case of "Phil Barker" or "The JISC Repositories
Research Team".
So I think the model as it stands probably needs two distinct entity
types classes, Agent and Agent Class (or something like that), in place
of the current Agent.
However, that's a provisional conclusion as I think this problem is
related to my next point.
(3) Context
I confess that whenever I see "Context" appear in a data model like
this, I come over a bit queasy :-) It is typically a very slippery
notion to define well, and sometimes once it is defined it turns out to
be a complex notion that needs to be modelled as a set of relationships
between several different things.
So I guess really my point is just that I think we need a better
description of what we mean by "context", and what specific operations
we are trying to support - really the latter rather than the former, I
think - before we can decide how best to represent it in an E-R model.
Maybe I should just stop there...but FWIW, it seems to me like there are
several different notions touched on in the document which are related
to "context"
(a) "intended audience". If, as suggested above, the object of an
is-Intended-For relationship is an Agent Class, rather than an Agent,
isn't it sufficient to say
(Learning Resource) XYZ is-Intended-For (Agent Class) Students in HE
What does the additional notion of a relationship between the Agent
Class and a "Context" bring to this case?
(b) _If_ there is something additional to be expressed (and I'm not
convinced yet that there is!), isn't it specific to the relationship
between resource and Agent Class i.e. won't you want to "say"
(Learning Resource) XYZ is-Intended-For (Agent Class) Students in HE
in-Context (Context) Some Context
(Learning Resource) PQR is-Intended-For (Agent Class) Students in HE
in-Context (Context) Another Context
If that is the case then I don't think that can be modelled as simply as
a binary relationship between the Agent Class and the Context; I think
it's a ternary relationship between Learning Resource, Agent Class and
the Context (which in a DCAM/RDF framework would have to be mapped into
a set of binary relations)
(c) I'm not sure whether the implication is that Agents (Persons,
Organisations), as well as Agent Classes, may be the subject of
"in-Context" relationships too, but the current model does suggest that.
What "context" is being captured here? Is that "affiliation" e.g. the
fact that the person called "Phil Barker" is affilated with the
institution called "Heriot-Watt University"? If that is the case, again,
I don't think that can be modelled as simply as a binary relationship
between the Agent and the Context. If I am the creator of two resources
at two different points in time, I may have two different affiliations,
so again we're probably dealing with a more complex n-ary relationship.
But again, I'm drifting into the realms of speculation. :-)
(d) My interpretation of the Educational.Context element in LOM is that
it simply expresses a relationship between a Learning Resource and an
"environment" and it doesn't involve an Agent at all.
In short, I think what is really needed is a clearer specification of
what operations/functions the model is seeking to support with what is
currently modelled as the in-Context relationship and the Context entity
type.
(4) Whole-Part Relationships
FRBR incorporates quite a wide range of Whole-Part relationships,
between instances of all the main FRBR entity types i.e. Work-Work,
Expression-Expression, Manifestation-Manifestation, Item-Item. For the
most part, SWAP sidestepped all of those relationships, I think, on the
grounds that they weren't required to support the basic functions
identified by the project.
I noticed that they had been omitted here too, and I just wanted to
double-check whether that was OK, or whether in fact there are
functions/operations which will require the modelling of Whole-Part
relationships. I'm almost reluctant to raise it as an issue as I
appreciate it requires some careful analysis of what FRBR entity types
are involved and it can make things quite complex! :-)
Cheers
Pete
[1]
http://efoundations.typepad.com/efoundations/2008/01/learning-materi.htm
l
---
Pete Johnston
Technical Researcher, Eduserv Foundation
Web: http://www.eduserv.org.uk/foundation/people/petejohnston/
Weblog: http://efoundations.typepad.com/efoundations/
Email: [log in to unmask]
Tel: +44 (0)1225 474323
|