zhengj2007 / bfo-trunk

0 stars 0 forks source link

domain and range of 'is preceded by'/'precedes' too narrow? #14

Open zhengj2007 opened 9 years ago

zhengj2007 commented 9 years ago

From dosu...@gmail.com on June 23, 2011 07:13:46

domain and range of 'is preceded by', precedes and their child properties is currently process. But shouldn't this be occurrent?

Here're the current definition:

occurrent: An entity that has temporal parts and that happens, unfolds or develops through time. Sometimes also called perdurants.

process: A processual entity that is a maximally connected spatiotemporal whole and has bona fide beginnings and endings corresponding to real discontinuities. SubClassOf: occurrent


A process must have 'bona fide beginnings and endings corresponding to real discontinuities', but if i subdivide a process or any other occurrent into fiat temporal parts, surely they will stand in 'is preceded by' relations to each other.

Original issue: http://code.google.com/p/bfo/issues/detail?id=15

zhengj2007 commented 9 years ago

From dosu...@gmail.com on June 28, 2011 08:20:59

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

zhengj2007 commented 9 years ago

From dosu...@gmail.com on June 28, 2011 08:22:59

Clairification:

Also - I'm posting this on the BFO google code tracker https://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

zhengj2007 commented 9 years ago

From dosu...@gmail.com on June 28, 2011 08:34:44

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

zhengj2007 commented 9 years ago

From pierregr...@gmail.com on June 28, 2011 09:07:12

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.

zhengj2007 commented 9 years ago

From janna.ha...@gmail.com on April 29, 2012 02:06:26

Domain and range for precedes and its inverse, is preceded by, are now `occurrent' in the current working version of BFO 2.

Status: Done
Owner: janna.ha...@gmail.com

zhengj2007 commented 9 years ago

From alanruttenberg@gmail.com on April 29, 2012 16:24:56

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

Labels: Resolution-proposal

zhengj2007 commented 9 years ago

From alanruttenberg@gmail.com on April 29, 2012 16:24:56

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

zhengj2007 commented 9 years ago

From alanruttenberg@gmail.com on April 29, 2012 16:25:33

Status: Accepted

zhengj2007 commented 9 years ago

From alanruttenberg@gmail.com on April 29, 2012 17:14:47

Raised issue regarding BFO2 reference as it doesn't currently define this relation. https://groups.google.com/forum/?fromgroups#!topic/bfo-devel/xlkp7_mHq8c

zhengj2007 commented 9 years ago

From dosu...@gmail.com on April 30, 2012 02:29:06

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'.

zhengj2007 commented 9 years ago

From alanruttenberg@gmail.com on April 30, 2012 05:14:59

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.

zhengj2007 commented 9 years ago

From dosu...@gmail.com on April 30, 2012 06:26:34

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.

zhengj2007 commented 9 years ago

From haen...@ohsu.edu on April 30, 2012 06:51:18

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.

zhengj2007 commented 9 years ago

From alanruttenberg@gmail.com on April 30, 2012 10:46:20

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.

zhengj2007 commented 9 years ago

From alanruttenberg@gmail.com on May 01, 2012 01:12:21

Labels: -Type-Defect Type-BFO2-Reference

zhengj2007 commented 9 years ago

From dosu...@gmail.com on May 02, 2012 08:55:04

"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?

zhengj2007 commented 9 years ago

From dosu...@gmail.com on May 02, 2012 08:55:07

"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?

zhengj2007 commented 9 years ago

From alanruttenberg@gmail.com on May 02, 2012 22:05:09

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.

zhengj2007 commented 9 years ago

From alanruttenberg@gmail.com on May 07, 2012 21:11:37

Owner: ifo...@gmail.com

zhengj2007 commented 9 years ago

From alanruttenberg@gmail.com on April 23, 2013 13:57:32

Labels: Milestone-BFO2-Release

zhengj2007 commented 9 years ago

From alanruttenberg@gmail.com on April 23, 2013 14:11:03

Blocking: bfo:164

zhengj2007 commented 9 years ago

From alanruttenberg@gmail.com on May 07, 2013 14:56:01

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.

zhengj2007 commented 9 years ago

From dosu...@gmail.com on May 09, 2013 02:45:46

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 here http://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.

zhengj2007 commented 9 years ago

From alanruttenberg@gmail.com on May 09, 2013 04:12:52

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.

Owner: alanruttenberg@gmail.com