Open GoogleCodeExporter opened 9 years ago
Pierre Grenon replied on BFO discuss:
Selected post Jun 23, 2011; 12:22pm
Re: Issue 15 in bfo: domain and range of 'is preceded by'/'precedes' too narrow?
Pierre Grenononline
this may hinge on a difference between a primitive relation and a
defined relation
in an axiomatisation of BFO, temporal order is primitive for time and
there are straightforward, almost trivial, and standard definitional
extensions for any temporal entity (e.g. if two occur at particular
times, the temporal order between the occurring entities follows that
between the corresponding temporal regions)
its credible that in an OWL encoding you would want to have vocabulary
that is notionally defined but for which the encoding contains no
definition
Original comment by dosu...@gmail.com
on 28 Jun 2011 at 3:20
Clairification:
- this comment refers to the version of BFO found using
http://purl.obolibrary.obo/bfo.owl , which resolves to:
http://bfo.googlecode.com/svn/trunk/src/ontology/bfo2.owl
Older versions may be better in some respects, but do not have the Foundry
compliant BFO URIs I need.
Also - I'm posting this on the BFO google code tracker
http://code.google.com/p/bfo/issues/detail?id=15. I would prefer answers on
this tracker ticket rather than on BFO discuss - or whatever other mailing
lists this might be forwarded to.
Cheers,
David O-S
Original comment by dosu...@gmail.com
on 28 Jun 2011 at 3:22
Hi Pierre,
Thanks for your reply. I'm afraid I'm rather struggling to follow it. The
problem I have is a simple practical one with the domain and range defined for
this object property: http://purl.obolibrary.org/obo/BFO_0000063 (may not
resolve - probably better to browse to it in Protege after loading
http://purl.obolibrary.obo/bfo.owl). I don't understand why domain and range
are specified as process, rather than occurrent. This choice has practical
implications for my ontology of Drosophila development, for which I would like
to use this object property. I believe that the relation/object property I am
currently using is equivalent to the BFO one, but I use it between stages.
Following the BFO definition for process and ocurrent, I classify stages as
occurrents rather than processes as stages may have fiat temporal boundaries.
Hope this is now clearer,
David
Original comment by dosu...@gmail.com
on 28 Jun 2011 at 3:34
hiya, sorry for the admittedly cryptic first reply, let me try again
I think what you experience is a need for a defined term, but it is not
something that motivates modifying primitive vocabulary. In my understanding,
the file you are accessing and using is a snapshot of a proposed reference
encoding of the basic (!) BFO (BFO2 at that). So it should contain primitive
vocabulary but not all sorts of possible definitional extensions.
Procedurally, I would argue this file should be kept minimal and if need be an
extended, distinct version with useful extensions be maintained concurrently.
It is likely such an extended version would prove more useful to many but it is
a different sort of work to find the right balance. This being said there are
obvious candidates for straightforward extension, like the one you mention.
Meanwhile, the quick fix would be to just make this extension in your ontology:
add a super property to BFO2's 'precedes', call it something like 'my generic
precedes', and set domain and range on occurrent. You could even set them on
entity, but it'll be a bit tricky to make sense of that and could be trickier
to axiomatise in OWL.
Original comment by pierregr...@gmail.com
on 28 Jun 2011 at 4:07
Domain and range for precedes and its inverse, is preceded by, are now
`occurrent' in the current working version of BFO 2.
Original comment by janna.ha...@gmail.com
on 29 Apr 2012 at 9:06
In BFO2 the distinction between processes, processual entities, and fiat
process parts, so David's objection is moot. The current proposed definition
(which I object to because of the use of s-depends) is "Definition: p is a
process = Def. a is an occurrent that has temporal proper parts and s-depends
on one or more material entities. [083-001]".
However resolution of my objection would not change this matter.
The problem with widening the domain and range to occurrent is that it
sanctions precedes relations between, for example, a spatiotemporal region
(which is an occurrent) and a process. That doesn't make any sense at the
moment, and to make it have sense would take work. If the only objection to
process as domain and range is based on David's argument, there is no reason to
change anything. Otherwise a new issue should be raised with the further
objection. I propose this issue be closed with the resolution that the domain
and range be process.
Note: There is no single current working version of BFO2. Janna's remark refers
to the version
https://bfo.googlecode.com/svn/trunk/src/ontology/owl-schulz/bfo.owl r207
Original comment by alanruttenberg@gmail.com
on 29 Apr 2012 at 11:24
In BFO2 the distinction between processes, processual entities, and fiat
process parts, so David's objection is moot. The current proposed definition
(which I object to because of the use of s-depends) is "Definition: p is a
process = Def. a is an occurrent that has temporal proper parts and s-depends
on one or more material entities. [083-001]".
However resolution of my objection would not change this matter.
The problem with widening the domain and range to occurrent is that it
sanctions precedes relations between, for example, a spatiotemporal region
(which is an occurrent) and a process. That doesn't make any sense at the
moment, and to make it have sense would take work. If the only objection to
process as domain and range is based on David's argument, there is no reason to
change anything. Otherwise a new issue should be raised with the further
objection. I propose this issue be closed with the resolution that the domain
and range be process.
Note: There is no single current working version of BFO2. Janna's remark refers
to the version
https://bfo.googlecode.com/svn/trunk/src/ontology/owl-schulz/bfo.owl r207
Original comment by alanruttenberg@gmail.com
on 29 Apr 2012 at 11:24
Original comment by alanruttenberg@gmail.com
on 29 Apr 2012 at 11:25
Raised issue regarding BFO2 reference as it doesn't currently define this
relation.
https://groups.google.com/forum/?fromgroups#!topic/bfo-devel/xlkp7_mHq8c
Original comment by alanruttenberg@gmail.com
on 30 Apr 2012 at 12:14
[deleted comment]
Fair enough. I brought this up because I've been treating stage as a sibling
of process.
Here's the definition of process that I took from Alan's OWL file and is still
present in Stefan's:
process: A processual entity that is a maximally connected spatiotemporal
whole and has bona fide beginnings and endings corresponding to real
discontinuities.
A stage is any temporal subdivision of a process that does not have 'bona fide
beginnings and endings corresponding to real discontinuities.'
The reason I brought this up is that I use preceded_by extensively (and
v.usefully!) between developmental stage terms. Developmental stages are
temporal slices of developmental processes (embryogenesis, oogenesis, wing disc
development). Their beginnings and endings are defined by fiat - based on the
timing of appearance or disappearance of easily score-able structures. So,
developmental stages are not, by the above definition, processes.
I would prefer a solution that keeps stage as a disjoint sibling of process, as
I think this mirrors that way biologists think about it and so will be useful
for error checking. Given that, we need to broaden the range of preceded_by,
even if not to 'occurrent'.
Original comment by dosu...@gmail.com
on 30 Apr 2012 at 9:29
Hi David,
It's not surprising that owl-ruttenberg has a stale definition - see above and
many other advertisements. But that definition is apparently no longer on the
table. There are a couple of points to make
a) We obviously need to understand and then work with a new version of the
definition of process. Coming to that definition is not a matter for the OWL
effort - it is an issue with BFO2 Reference. Note that the proposed BFO2
reference does not have a term 'processual entity'.
b) If you have no contest to the current definition, then I'd offer that this
isn't a BFO issue. Rather you could, if desired, define two subclasses of
process corresponding to (I guess) the whole of an organism's life, and a stage
of their life (proper temporal part of the whole). It is unclear whether, for
your purposes, would need any different definition of preceded_by than is
defined for BFO, but that's a matter for you (developmental anatomists) to sort
through.
Original comment by alanruttenberg@gmail.com
on 30 Apr 2012 at 12:14
Hi Alan,
Taking it outside of BFO might work. But first I want to check -
Given:
Definition: p is a process = Def. p is an occurrent that has temporal proper
parts and s-depends on one or more material entities. [083-001]
Would a process defined in this way encompass developmental processes like
embryogenesis, oogenesis and 'wing disc metamorphosis' and also every possible
temporal slice of these processes?
If this is the case, then I would still want at least 'developmental process'
and 'developmental stage' as disjoint sibling SubClasses of process, defined
along the lines I've already discussed. But it surprises me that others would
not be interested in this as a general distinction.
Original comment by dosu...@gmail.com
on 30 Apr 2012 at 1:26
I am not up to date on this discussion, but I wholeheartedly agree with David
that we need a precedes relation between occurrents that have fiat boundaries
to represent stages. This cannot be in an extended file or a quick fix - if we
don't have this in a core BFO release, the anatomical ontologies will not use
BFO (and in fact most do not now for this and other reasons).
Is there a requirements document outlining the anatomy needs? It would seem
this would be helpful to ensure they are met.
Original comment by haen...@ohsu.edu
on 30 Apr 2012 at 1:51
Hi David,
As I said, I don't like the current BFO2 Reference definition. One of the
reasons is that I have no way of formally evaluating your question, because I
don't think an s-depends relation between occurrents and continuants makes any
sense. Second, I don't like having (or not) temporal parts as a differentia
because being a temporal part seems more of a formal ontology sort of thing
rather than a physically grounded kind of thing.
That said, I think it is the intention that the term subsume your cases.
Those classes, then, should, IMO, go in some more specific ontology (perhaps
CARO? GO?) .
Regarding fiat and bona fide boundaries of processes, I don't know how to
determine this distinction on process boundaries, so I personally wouldn't want
there to be such a distinction. I think the right abstraction has to do with
whole organismal lives versus proper parts of them.
Melissa, the proposal isn't to put preceded_by in a separate file, but rather
that if David wants the subclasses 'developmental process' and 'developmental
stage' then those should live not in BFO, as they are too specific for BFO. The
proposal I made (leave domain/range of process as process) is compatible with
that scenario.
Original comment by alanruttenberg@gmail.com
on 30 Apr 2012 at 5:46
Original comment by alanruttenberg@gmail.com
on 1 May 2012 at 8:12
"Regarding fiat and bona fide boundaries of processes, I don't know how to
determine this distinction on process boundaries, so I personally wouldn't want
there to be such a distinction. I think the right abstraction has to do with
whole organismal lives versus proper parts of them."
(i) As mentioned in previous comments, stages can apply to developmental
processes as well as to the process that is development of the whole animal.
(ii) Do you believe that processes that end have no bona-fide temporal boundary
that is there end. e.g.- about 42 years ago, I gastrulated. I am not
gastrulating right now. Would you maintain that the point in time at which I
stopped gastrulating can only be defined by fiat. Is the temporal boundary that
is the end of me gastrulating a different sort of temporal boundary during to
the one I declare by fiat to have occurred 5 minutes after this instance of
gastrulation started?
Original comment by dosu...@gmail.com
on 2 May 2012 at 3:55
"Regarding fiat and bona fide boundaries of processes, I don't know how to
determine this distinction on process boundaries, so I personally wouldn't want
there to be such a distinction. I think the right abstraction has to do with
whole organismal lives versus proper parts of them."
(i) As mentioned in previous comments, stages can apply to developmental
processes as well as to the process that is development of the whole animal.
(ii) Do you believe that processes that end have no bona-fide temporal boundary
that is there end. e.g.- about 42 years ago, I gastrulated. I am not
gastrulating right now. Would you maintain that the point in time at which I
stopped gastrulating can only be defined by fiat. Is the temporal boundary that
is the end of me gastrulating a different sort of temporal boundary during to
the one I declare by fiat to have occurred 5 minutes after this instance of
gastrulation started?
Original comment by dosu...@gmail.com
on 2 May 2012 at 3:55
David asks: "Do you believe that processes that end have no bona-fide temporal
boundary that is there end. e.g.- about 42 years ago, I gastrulated. I am not
gastrulating right now. Would you maintain that the point in time at which I
stopped gastrulating can only be defined by fiat. Is the temporal boundary that
is the end of me gastrulating a different sort of temporal boundary during to
the one I declare by fiat to have occurred 5 minutes after this instance of
gastrulation started?"
I can recognize when the 3-layered structure has started to form, and when it
seems complete. If the temporal projection of the process is defined to end
when that occurs then we have some information about the process.
First, remember that a process boundary is a spatiotemporal boundary. So it
isn't enough to say the time to make a process boundary. But first look at the
temporal boundary. When is the end of gastrulation? When the last cell
necessary to complete the structure first starts to exist? When it falls into
some position (which)? Just before the next cell division?
As for spatial extent, what is it? Is it the sum of the spatial regions of the
cells? Are there not metabolites with various functions surrounding the
anatomical structure? How far out do these go? How much of that contributes to
the spatial extent of the process.
I think there are enough questions to merit my assessment that, ontologically,
making these distinctions is no simple matter.
GO:0007369
Definition
A complex and coordinated series of cellular movements that occurs at the end
of cleavage during embryonic development of most animals. The details of
gastrulation vary from species to species, but usually result in the formation
of the three primary germ layers, ectoderm, mesoderm and endoderm.
Original comment by alanruttenberg@gmail.com
on 3 May 2012 at 5:05
Original comment by alanruttenberg@gmail.com
on 8 May 2012 at 4:11
Original comment by alanruttenberg@gmail.com
on 23 Apr 2013 at 8:57
Original comment by alanruttenberg@gmail.com
on 23 Apr 2013 at 9:11
David, we discussed this on the BFO call today. Do you feel that defining
precedes as narrowly (but consistently with RO2005) having domain and range
process (BFO2 sense which doesn't distinguish between fiat and bona fide
process boundaries) will serve your current purposes and not block you? If so,
the preference is to keep it as is and address and other needs for relations
among occurrents for the next release.
Chris, has RO currently defined precedes/immediately precedes relations? If so
I would advocate moving them to BFO but not changing their IDs. If not I will
allocate new ids.
Original comment by alanruttenberg@gmail.com
on 7 May 2013 at 9:56
OK. In the interests of progress, I'm happy to for the domain and range of
preceded_by / precedes to be BFO:process as currently defined. I do believe
there is an important distinction between processes, in the generally
understood sense of the term, and arbitrary temporal slices of them, which is
what stages are (*further details below in case you are interested). We will
still need to find a home for two disjoint sibling subclasses: biological stage
and biological process.
As to the relations: I think it makes sense to use Allen relations and perhaps
their derivatives. Although I worry that the domain and range restrictions are
too strong for this - wouldn't it be valid to use Allen relations between
individual one-dimensional temporal regions?
John Goodwin has made a well axiomatised OWL version of Allen - including the
composition table expressed as property chains (see:
http://johngoodwin225.wordpress.com/2012/04/22/about-time/) I see no reason
to replicate this work. We should just ask him if he'd be happy to request an
obolibrary IDspace and generate PURL URIs for the terms. I guess whether they
end up in BFO will depend on whether we agree that BFO:process is a valid D/R
for Allen relations. If not, we should simply script making sub-properties for
BFO with the D/R restrictions added.
I've made a start at mapping the timing relations we've been using to Allen.
See:
https://docs.google.com/document/d/1kBv1ep_9g3sTR-SD3jqzFqhuwo9TPNF-l-9fUDbO6rM/
edit
Some of them are actually disjunctive combinations of Allen relations-
happens_during corresponds to Allen 'sfo' (following the nomenclature
herehttp://www.thomasalspaugh.org/pub/fnd/allen.htm ) - while others are vaguer
than Allen (starts_during). Some combined relations are, I think, more
intuitive than the Allan relations so are probably worth keeping. This choice
will have to be balanced against any drawbacks of using the combined relations.
This needs more discussion, but should not hold up importing the Allen
relations themselves.
-------------
*Some further discussion of stages vs 'processes'.
We may not have found the perfect differentium, but I think it is perfectly
reasonable to distinguish (as most biologists would) between processes such as
pupariation, gastrulation or cell division, and arbitrary temporal subdivisions
of them. Stages are the latter. They are simply standard systems for
temporally subdividing biological processes. Developmental stages in particular
are typically defined for purely pragmatic reasons. For example, the end of
the first stage of gastrulation in Drosophila melanogaster is defined as the
time at which the posterior plate is in the same plane as the
anterior-posterior axis. Unlike gastrulation itself, there's nothing
particularly special about this event or how it divides gastrulation - no
reason to believe that we might make generalisations across species about it,
or the slice of gastrulation that precedes it. It is simply easy to score.
Original comment by dosu...@gmail.com
on 9 May 2013 at 9:45
OK, thanks David.
I think the Allen relations make sense for temporal intervals (over which they
are typically defined, but can be extended to processes using the 'projects on
to temporal region' relation from processes to temporal regions. However, Note
Bijan's comments about performance/completeness - FOL/DL isn't an idea system
for doing the actual Allen calculation, which is typically done using
constraint programming. The wikipedia page for the allen relations points to a
java implementation. In DL land, it would be curious to see if it could be
implemented as a concrete domain. I'd suggest talking with Bijan for practical
strategies. Barry has also indicated that he would like the Allen relations as
part of BFO. My only concern is performance/completeness/expectation setting.
I'm going to close this as resolved once I add in the precedes relations but
start a new issue for the Allen relations, making you owner.
Original comment by alanruttenberg@gmail.com
on 9 May 2013 at 11:12
Original issue reported on code.google.com by
dosu...@gmail.com
on 23 Jun 2011 at 11:13