Current view of 'anatomical location'?

[Thomas Beale]

This is an old question, but since I believe there is always ongoing thinking in the area of BFO:Site, Immaterial Entity, continuant boundaries, and suchlike, I thought I'd ask today, to see if there is any new thinking on the question: what is an 'anatomical location'?

NB: I am mainly thinking of derivative informations models rather than BFO itself, so this is a 'using BFO' question, not a 'what's wrong/right with BFO' question...

A previous thread on this is to be found here (where Barry posted the FMA Anatomical space / structure graphic), but with no super-precise answer. I am aware of voluminous prior discussions of course, but have only perused a few.

This is asked from a practical point of view of whether the answer filling a typical field like 'body location' in a clinical data system can be said to be an instance (or ref to) of a single ontological class.

The practical problem is that such answers can easily identify either a space (i.e.a BFO Immaterial entity) or a structure (a BFO Material entity) - they're both possible sites of some clinical observation or activity.

Mentally I tend to see the identification of a site, including a structure (such as a particular finger etc.) as identifying a 2- or 3D continuant fiat boundary, i.e. a space, rather than the physical structure or cavity enclosed by the space per se. This does not seem that unreasonable since the meaning is usually that something will happen somewhere in the identified space, e.g. a biopsy needle will be used on an identified lump in the left breast. We don't usually identify exactly what material place that needle will go on; we just refer to the lump as a whole.

If we were to literally follow BFO within the kinds of information models I am thinking of, the formal type for 'anatomical location' (in the sense described above) would be a disjunction of (approximately) Site and Material entity, which is messy.

[Thomas Beale]

[Small correction: there's no 3-D continuant fiat boundary; I meant just 2-D continuant fiat boundary.]

[Werner Ceusters]

'body-location' would correspond to BFO2020's range of the relation 'located-in' which can be filled by any independent-continuant (IC) which is not a spatial-region (SR) as per axiom:

  (cl:comment "located-in is time indexed and has domain: independent-continuant but not spatial-region and range: independent-continuant but not spatial-region [bge-1]"
    (forall (a b t)
     (if (located-in a b t)
      (and
       (and (instance-of a independent-continuant t)
        (not (instance-of a spatial-region t)))
       (and (instance-of b independent-continuant t)
        (not (instance-of b spatial-region t)))
       (instance-of t temporal-region t)))))

For 'body-location', the range should be such IC which is part of the (human, I suppose) body, and which includes sites such as the armpits, ear canal, etc.

Your needle biopsy example is still satisfied by BFO's located-in because of this axiom:

  (cl:comment "If something is located in something else then the region of the first is part of the region of the second [uas-1]"
    (forall (a b t)
     (if (located-in a b t)
      (exists (r1 r2 t2)
       (and (temporal-part-of t2 t) (occupies-spatial-region a r1 t2)
        (occupies-spatial-region b r2 t2)
        (continuant-part-of r1 r2 t2))))))

In no way would 'anatomical location' for BFO be a type. But you could use the term as shorthand for 'independent continuant which is part of the human body and which is stated to be the location of something'. The 'stated' in the definition makes clear it is not a definition of a type, but an explanation for what would fit in the field of your (perhaps messy :-) ) information model. 

W

[Thomas Beale]

Werner,

thanks for the reply. That is about what I expected. The conjunction 'independent-continuant (IC) which is not a spatial-region (SR)' is nearly equivalent to the disjunction 'Site or Material entity'.

> In no way would 'anatomical location' for BFO be a type.

Certainly not in current BFO.

Although... the formula 'IC which is not a SR' is not uncommon in BFO due to spatial regions having no dependence on Material or Immaterial Entities, whereas Immaterial Entities are dependent on Material Entities for their physical form and location, although not existence (perhaps a notional 'p-depends' relation). I have often wondered if there were conversations long ago in which a cleaner separation was discussed, such that Spatial Regions were distinct from something called (say) Physical Entity. I.e. something like:

Independent Continuant

    Physical Entity

        Material Entity

        Immaterial Entity (p-depends on Material Entity)

    Spatial Region

In such a hierarchy, any Physical Entity could indeed be an 'Anatomical location', and the need for 'IC which is not a SR' would disappear.

- t

[Werner Ceusters]

The 'independent-continuant (IC) which is not a spatial-region (SR)' which you claim to be a conjunction, is taken out of the context in which it was used: you should not omit the 'any' which was just in front of it, and you should not omit the axiom which I put there to compensate for the economic - but ambiguous way - in which natural language works and by means of which I made that statement. Also do not forget that BFO is about faithful representation independent of any purpose. If you use a tool or language that cannot deal with 'not' or with time indexing,  or with explicit representation of individuals, then you cannot use that as an argument for inventing types to be put in BFO simply to make your work easier. But of course, you are free to create an application ontology 'ap' that has a class 'ap:Physical Entity' and another one 'ap:Immaterial Entity' (the latter distinct from 'bfo2020:immaterial entity' and define them as you see fit. You should however be careful in your axioms that you don't violate the ones for the bfo types you do include in ap. That is the danger of using for example OWL DL. It has recently been shown that OBI, one of the very few ontologies that tried to stay within BFO's view, contained inconsistencies which popped up by including the FOL axioms. Also, BFO might indeed evolve in the future and take another position on sites or holes in general (tricky issues, see https://plato.stanford.edu/entries/holes/). But for the time being, it uses parthood as a relation between sites and what they are bounded by:

  (cl:comment "If a has-continuant-part b then if a is an instance of site then b is an instance of site or continuant-fiat-boundary  [mjj-1]"
    (forall (p q t)
     (if (and (has-continuant-part p q t) (instance-of p site t))
      (or (instance-of q site t)
       (instance-of q continuant-fiat-boundary t)))))

  (cl:comment "If a continuant-part-of b then if a is an instance of site then b is an instance of site or material-entity  [izr-1]"
    (forall (p q t)
     (if (and (continuant-part-of p q t) (instance-of p site t))
      (or (instance-of q site t) (instance-of q material-entity t)))))

 
Alan Ruttenberg did a tremendous job in axiomatizing BFO. Perhaps it is not completely in the way me and you in the past (mis)understood parts of BFO, but it offers now a clear way to look at it. 

My suggestion is to base any discussion on what (perhaps) needs to be added or changed in BFO, by looking at the existing FOL axioms and demonstrating that there are unintended models.   

You might also want to read Alan Rector and colleague's excellent paper about misunderstandings about representation languages: https://pubmed.ncbi.nlm.nih.gov/34384571/ .

[Thomas Beale]

Agree with all that.

That is indeed a very good paper.

[Chris Mungall]

I support this. The expression IC and not an SR is an awkward construct. I know BFO lives in the realm of abstract philosophy and doesn't care too much for making things easier for users, but this proposed grouping really simplifies things.

I still don't know of any ontology that makes use of SR, which adds to the argument for a simple grouping that excludes it.

--
You received this message because you are subscribed to the Google Groups "BFO Discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to bfo-discuss...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/bfo-discuss/b19fa2cf-b81d-44a5-9590-e5115fc70a74n%40googlegroups.com.

[Thomas Beale]

A further rumination on the BFO taxonomy in this area. Just posting this as a logical completion of the previous post. Nothing here is intended as any kind of change request or complaint about BFO; it's just for interest, since I happen to be working on information models that relate.

The following shows what I think is a clear way to deal with structures and spaces in anatomy, objects/artefacts and geographical / environmental. It seems a bit cleaner than BFO, and I think could be axiomatised in a way very close to current BFO but with obvious adjustments.

    Material entity

        Structure - substantial entity whose physical structure independently persists in time (approximately solid state); may contain Material entity/ies; == BFO:Material entity

        Space - cavity created by a Structure(s) (may contain gas, plasma, radiation, nothing) - causally dependent on containing Structure; == BFO: Site

    Immaterial entity

        Materially dependant locus - non-substantial entity defined in relation to a Material entity host, of 1 dim less than the host; == BFO: Continuant fiat boundary

            Material point - point whose location is relative to some part of a Material entity; == BFO: 0-D Continuant fiat boundary

            Material boundary - linear (1D)  on a surface of some part or whole Material entity; == BFO: 1-D Continuant fiat boundary

            Material surface - bounding surface (2D) of some part or whole 3D Material entity; == BFO: 2-D Continuant fiat boundary

        Materially independent spatial locus - non-substantial entity defined with respect to a frame-of-reference, which may be located relative to a Material entity; == BFO:Spatial region

            Spatial point - linear boundary on a surface of some part or whole Material entity; == BFO: 0-D Spatial region

            Spatial boundary - linear boundary on a surface of some part or whole Material entity; == BFO: 1-D Spatial region

            Spatial surface - bounding surface of some part or whole Material entity; == BFO: 2-D Spatial region

            Spatial volume - bounding surface of some part or whole Material entity; == BFO: 3-D Spatial region

The main change (linguistic and ontological) is not to treat cavities / spaces within structures as 'immaterial', but as substantial (even if in specific cases, a perfect void obtains). Thus, such things are distinct from truly immaterial entities which divide into either materially-dependent loci (boundaries on material things) and materially-independent loci (boundaries w.r.t. frame-of-reference that is independent of the material entities located by it).

Accordingly, there could be no causal relation between a Material entity and an Immaterial entity in the above, nor causal involvement of an Immaterial entity in anything. The way BFO stands today, I would suggest that the possibility of causal involvement of some bfo:Immaterial entities (Sites) but not others is problematic, and complicates both the axiomatisation and therefore inferencing (hence the annoying recurring formula 'IC which is not SR'). This doesn't mean it makes it wrong of course, but it's always nice to keep things as simple as possible.

[Werner Ceusters]

Since your proposal is not about BFO, but rather about an application ontology you seem to need and for which for some reason you want to relate to bits and pieces of BFO, it would be better to discuss it on an OBO discussion list. In OBO ontologies they seem not to be too fussy in the interest of simplicity about mixing linguistics with ontology, putting second-order reality prior to first-order (as in your description of materially dependent locus: '... defined in relation to ...'), using 'or' in definitions, making definitions on the basis of terms that are not defined or explained (like 'substantial entity' and 'cavity') or have labels which exhibit odd face value (like 'material point' being defined as not a point but something linear), etc. Ignoring many principles that BFO uses for ontology design makes it indeed easier and more simple to very quickly come up with representations of thousands of entities.

Nevertheless, I am puzzled about you having a problem with the expression 'IC which is NOT an SR', but not with your proposed term 'non-substantial entity'. Does that not mean 'whatever is NOT a substantial entity'? 

[Werner Ceusters]

In my previous post, I need to take the phrase "(like 'material point' being defined as not a point but something linear)" back. My old eyes got the lines about material point and material boundary blurred. Sorry for that.

[Thomas Beale]

> Since your proposal is not about BFO, but rather about an application ontology you seem to need and for which for some reason you want to relate to bits and pieces of BFO, it would be better to discuss it on an OBO discussion list.

In common with many projects and efforts, we use BFO as a reference in the first instance for thinking (and sometimes computing) about biomedical (mainly clinical and analytic) information. I've posted some notes here for interest, since they are specifically about BFO and not some other ontology.

> putting second-order reality prior to first-order (as in your description of materially dependent locus: '... defined in relation to ...'),

Not sure I get your objection here. That's just slightly different language for the description of bfo:Continuant fiat boundary found in the BFO Reference i.e. "Intuitively, note(continuant fiat boundary)[a continuant fiat boundary is a boundary of some material entity (for example: the plane separating the Northern and Southern hemispheres; the North Pole), or it is a boundary of some immaterial entity (for example of some portion of airspace)."

Additionally, the Editor note to BFO: Immaterial Entity (Ontobee) has '... Immaterial entities are divided into two subgroups:boundaries and sites, which bound, or are demarcated in relation, to material entities ...'

Without an underlying Material Entity or Immaterial Entity (but not, I think, Spatial region), a Continuant fiat boundary cannot exist - it is a boundary that separates some thing from some other thing.

> using 'or' in definitions

If you are referring to 'some part or whole Material entity' that is just some NL to make the definition intuitive. I am not trying to fully and precisely articulate an alternative ontology here (that would take some careful work). Indeed, I had not meant to capitalise 'Material' in those phrases, i.e. I meant to indicate just that we are talking about boundaries on material entities, or parts thereof, which formally just equates to 'Material entity'. I don't think anyone on this list will be confused by that...

> making definitions on the basis of terms that are not defined or explained (like 'substantial entity' and 'cavity')

Substantial = consisting of substance i.e. matter / energy; Cavity just means the same as bfo:Site, i.e. 'b is a site means: b is a three-dimensional immaterial entity that is (partially or wholly) bounded by a material entity or it is a three-dimensional immaterial part thereof.'

I used somewhat different language in the taxonomy so as not to confuse directly with BFO definitions.

> Nevertheless, I am puzzled about you having a problem with the expression 'IC which is NOT an SR', but not with your proposed term 'non-substantial entity'. Does that not mean 'whatever is NOT a substantial entity'?

This is the issue of interest. I argue that ontically, that a bfo:Site is not immaterial but material. Then boundaries fall out cleanly as being causally dependent on Material entities, including cavities, voids, holes etc. The rest follows.

I am interested to know what substantive BFO precepts are broken by this reasoning (maybe there are some...)

- t

[Werner Ceusters]

 WC:Since your proposal is not about BFO, but rather about an application ontology you seem to need and for which for some reason you want to relate to bits and pieces of BFO, it would be better to discuss it on an OBO discussion list.

TB: In common with many projects and efforts, we use BFO as a reference in the first instance for thinking (and sometimes computing) about biomedical (mainly clinical and analytic) information. I've posted some notes here for interest, since they are specifically about BFO and not some other ontology.

WC: You wrote in a previous reply: "Nothing here is intended as any kind of change request or complaint about BFO".

WC: putting second-order reality prior to first-order (as in your description of materially dependent locus: '... defined in relation to ...'),

TB: Not sure I get your objection here. That's just slightly different language for the description of bfo:Continuant fiat boundary found in the BFO Reference i.e. "Intuitively, note(continuant fiat boundary)[a continuant fiat boundary is a boundary of some material entity (for example: the plane separating the Northern and Southern hemispheres; the North Pole), or it is a boundary of some immaterial entity (for example of some portion of airspace)."

WC: entities need to exist before you can define them or relate them. And when you discover that a number of entities exist, then you can discover that they are related in a certain way. And then you write axioms to describe these relationships. That principle is different than defining classes on the basis of properties. BFO is not about classes or about throwing properties together - or separating them - because it is convenient for one or other purpose.

TB: Additionally, the Editor note to BFO: Immaterial Entity (Ontobee) has '... Immaterial entities are divided into two subgroups:boundaries and sites, which bound, or are demarcated in relation, to material entities ...'

WC: Yes, I don't see how that contradicts my point or justifies yours. The note first claims the existence of boundaries and sites, and then uses some language to state how they are related to material entities.

TB: Without an underlying Material Entity or Immaterial Entity (but not, I think, Spatial region), a Continuant fiat boundary cannot exist - it is a boundary that separates some thing from some other thing.

WC: Not really. You are understanding "boundary" not in the way BFO does. That is why I earlier said we should only discuss on the basis of the axioms. Look at the axioms in continuant-mereology.cl. and spatial.cl There is not a single one that relates a continuant-fiat-boundary to anything else than the one entity it is a continuant-part of (besides time of course).

TB > using 'or' in definitions

If you are referring to 'some part or whole Material entity' that is just some NL to make the definition intuitive. I am not trying to fully and precisely articulate an alternative ontology here (that would take some careful work). Indeed, I had not meant to capitalise 'Material' in those phrases, i.e. I meant to indicate just that we are talking about boundaries on material entities, or parts thereof, which formally just equates to 'Material entity'. I don't think anyone on this list will be confused by that...

WC: I am. But that is the point! The problem is language! And you are right, it takes careful work. BFO has after 20-something years 35-something types. You offer here in two days 8 new ones. Take your time and analyze them carefully. At BFO's rate, you are entitled to roughly 5 years !  :-)

WC: > making definitions on the basis of terms that are not defined or explained (like 'substantial entity' and 'cavity')

TB: Substantial = consisting of substance i.e. matter / energy; Cavity just means the same as bfo:Site, i.e. 'b is a site means: b is a three-dimensional immaterial entity that is (partially or wholly) bounded by a material entity or it is a three-dimensional immaterial part thereof.'

I used somewhat different language in the taxonomy so as not to confuse directly with BFO definitions.

WC: but it does not help.  

WC: > Nevertheless, I am puzzled about you having a problem with the expression 'IC which is NOT an SR', but not with your proposed term 'non-substantial entity'. Does that not mean 'whatever is NOT a substantial entity'?

TB: This is the issue of interest. I argue that ontically, that a bfo:Site is not immaterial but material. Then boundaries fall out cleanly as being causally dependent on Material entities, including cavities, voids, holes etc. The rest follows.

WC: Two things here. 

1) You ignore the point of the NOT that I made. Either you accept the use of "not", then the entire discussion was moot because "IC which is not SR" is fine. Or you don't, but then you cannot have "non-substantial".

2) If holes are substantial, like you propose, then by your definition they consist of matter / energy. How much of the weight of a piece of  Emmentaler cheese on Earth is contributed to by the holes then, or what amount of the energy you receive by eating a piece is contributed to by the holes?  If you drop a glass and it brakes in pieces on your floor, how much time are you going to spend to look for pieces of the material of which the glass' cavity consisted so you will not hurt yourself later by walking barefoot there?

TB: I am interested to know what substantive BFO precepts are broken by this reasoning (maybe there are some...)

WC: well, I addressed them before and here.

But to conclude this, I can't strengthen enough the importance of the axioms. Alan Ruttenberg has explained in several contributions to other conversations in this google-group how the DL-version is but an approximation of BFO constructed on the basis of what can be generated from the axioms, and what can be expressed because of the DL limitations. I prefer to call it a 'crippled' version. I understand it has its uses - so have guns and drugs and prostitutes - but one should not discuss BFO on the basis of what is in the OWL version. 

 

- t

[Thomas Beale]

WC: 1) You ignore the point of the NOT that I made. Either you accept the use of "not", then the entire discussion was moot because "IC which is not SR" is fine. Or you don't, but then you cannot have "non-substantial".

you're right, sorry about that, I read past it too fast. I'll come back to this point. BTW I have reviewed the formal axioms (for others, the location is here on GitHub); I'm not trying to avoid those. (It is impressive work, indeed.)

On the metaphysical question of Sites (lumen of gut, holes in swiss cheese, the grand canyon...) - these are all some kind of cavity created by a Material solid structure and/or fiat demarcation (the airspace example) related to Material structure(s). Physically speaking, such places contain matter (air, dust, liquid, gel...) and/or energy (radiation within some otherwise bit of clean space considered 'part' of the International Space Station). In the degenerate case, someone might be able to find a site containing absolutely nothing at all.

In all these cases, the so-called site is physically and causally part of the containing Material entity. We might be very interested in what the 'site', e.g. what gas, radiation, pollutants... how about a floating airship, or soap bubbles? There's a kind of continuant part-hood that applies among parts of any physical entity, including all its 'sites'. If that is not what BFO means by 'site', then the only meaning can be a pure 3-D geometric volume that corresponds to the cavity / hole / etc, i.e. the 3D space occupied by the physical contents of the site - a true 'immaterial' entity. Or perhaps a 2-D bounding surface? But then it would be a 2-D continuant fiat boundary. If it really is a non-physical thing, then it has no existence other than as information (in a computer system, on paper etc). A different conception of parthood is surely needed between truly material and truly immaterial entities than exists between material entities.

Feel free to correct this understanding of Site - but note, I think we should see the axioms as following and formalising the metaphysical explanation, not logically prior to it.

- t

[Werner Ceusters]

If you look at all axioms in all .cl files for site, then you see there is nothing fishy and everything corresponds nicely to what I said before. I have been working with them for over a year now and I am trying to develop for educational purposes in my courses a suite of reasoners that work on what can be generated for different purposes directly out of the axioms.

The BFO2.0 specs provide elucidations and definitions for what material entity, immaterial entity, site, etc mean. 

a(material entity)[Elucidation: A material entity is an independent continuant that has some portion of matter as proper or improper continuant part. [019-002]]

a(immaterial entity)[Definition: a is an immaterial entity = Def. a is an independent continuant that has no material entities as parts. [028-001]]

a(site)[Elucidation: b is a site means: b is a three-dimensional immaterial entity that is (partially or wholly) bounded by a material entity or it is a three-dimensional immaterial part thereof. [034-002]

For the metaphysics behind them, one can read:  :Immaterial entities are independent continuants which contain no material entities as parts. The roots of BFO’s treatment of such entities lie in the application of theories of qualitative spatial reasoning to the geospatial world, for example as outlined in [49, 69], in the treatment of holes by Casati and Varzi [48], in the treatment of niches by Smith and Varzi [7, 10] and in the treatment of cavities in the FMA [43, 44, 34, 35]."

A problem with the above eluc/defs is that they are not time indexed

I do have a document that says BFO2020 specifications and that dates from more than a year back. I can't find it online and don't know what its current status is. I don't even know whether it is part of the standard, or work in progress.

Therein I find:

A.1.2.17 material entity

Elucidation: A material entity is an independent continuant that at all times at which it exists has some portion of matter as continuant part [019-BFO]

and:

A.1.2.23 immaterial entity

Definition: a is an immaterial entity =Def. a is an independent continuant which is such that there is no time t when it has a material entity as continuant part at t [028-BFO]

A.1.2.28 site

Elucidation: b is a site means: b is either (1) a three-dimensional immaterial entity that whose boundaries either (1) (partially or wholly) coincide with the boundaries of one or more material entities or (2) have locations determined in relation to some material entity [034-BFO]

There is something grammatically wrong with that def, but the intend seems clear.

None of these defs contradict or are contradicted by any of the axioms. When parthood therein is used, then it is such that material entities can have immaterial ones as part, but not the other way round.

What is also specified is this: The above elucidations of site and of the different kinds of continuant fiat boundary will be replaced by definitions when dimension and boundary dependence have been defined within the BFO framework.

Maybe that is what you had in mind with "p-dependece" earlier? Even so, I still don't see what it has to do with sites being filled or not. Whatever is in a site is located in it, not part of it. Whereas boundary dependence probably would be in relation to the 'host', it would still be independent of site content.

W

[dl...@mitre.org]

Werner et al;

 "when dimension and boundary dependence have been defined within the BFO ". I'm very interested in what these might be. I have a mental place-holder "b-depends" relation that's sort of based on Barry's "Drawing Lines on a Map" paper from way back.  Anything available on what these might look like?

thanks

  BTW, I think the document you referred to was a version of "Natural language specification of BFO 2020" doc that used to be on the http://basic-formal-ontology.org/ site. Removed for some reason.

[Werner Ceusters]

I have nothing on "b-depends" or "p-depends" on my laptop which is thoroughly indexed and contains all my stuff collected since 1991. I don't remember discussions about it that I might have participated in. Sorry.

[Thomas Beale]

Initial reaction...

The definition of Site is such that it contains no matter (following the axioms). Since real cavities, holes in cheese, space in a ship's hold etc all contain matter (and may do so in complicated ways, e.g. a sparrow flying in a ship's hold), a Site can only be understood as a geometric specification that has no real-world counterpart, i.e. a 3-D volume (the enclosing 2-D bounding surface would be just as good). It is therefore purely informational and can be represented somewhere, but it does not physically exist. But since that description is also no different from BFO:2-D continuant fiat boundary (other than the quibble about 2-D or 3-D), it can't be that.

How then do the has-continuant-part relations in the axioms obtain between Material things and Sites (or any Immaterial things)? The sense in which parthood obtains between a solid ship and its (air- or grain-filled) hold cannot be the same sense as between the ship and the Site representing the hold. The fact of a sparrow flying in the hold would seem to complicate things: if the hold is Immaterial, it cannot contain any Material entity at any time. The ship's hold understood as a BFO:site cannot possibly contain a bird, and yet in the real world it does.

I believe you will say: no, the bird is only located in the hold (Site). But the bird is physically located within a material entity, namely the air space of the ship's hold. Its location (but not it) is inside the volumetric shape (or boundary surface) of the Site, as an abstract geometric relation. (I think, between a 0-D Continuant fiat boundary (a point) corresponding to the Quality of the bird's location, and a closed 2-D Continuant fiat boundary)

(There might be a possible argument that says a bird is clearly not 'part-of' any ship in the sense of intentional design, but this fails with time indexing. A ship whose hold contains a flying bird for some time certainly has-continuant-part bird for that time. If it is destroyed by a missile, the ship and all its 'parts', including the bird, will be destroyed. The bird is part of the transitive closure of the ship while it is inside it.)

Some of these issues are touched upon in the introduction of Varzi's SEP Mereology article.

It seems to me that solutions to the above require different types of 'parthood' to be distinguished, at least to allow:

• material containment, and
  
• boundaries located in/on material hosts

to be distinct.

- t

[Thomas Beale]

Just to get back to the point I missed earlier:

WC: 1) You ignore the point of the NOT that I made. Either you accept the use of "not", then the entire discussion was moot because "IC which is not SR" is fine. Or you don't, but then you cannot have "non-substantial".

The problem is not the negation; the problem is that 2 sub-categories of Immaterial entity (Site and Continuant fiat boundary) are dependent upon Material hosts (I am avoiding the word 'bearer' here...) whereas the 3rd (Spatial region) is independent of any Material entity. This obviously matters because the formula 'IC which is not SR' recurs throughout BFO, including its axioms, and it is obvious why: every Independent Continuant has a Materially related host, i.e. is a Material entity or a part thereof (including Immaterial sub-parts) - except Spatial Regions, which have nothing to do with 'things', they are just parts of a spatial reference frame.

I am almost tempted to say that IC should have as its sub -categories Materially-dependent and Materially-independent. 'IC which is not an SR' would then become just Materially-dependent.

My term 'non-substantial' was meant metaphysically, not axiomatically, i.e. meaning 'having no substance' (matter / energy).

- t

[Werner Ceusters]

Like I said: material entities can have per BFO sites and boundaries as part, but not the other way round.

The bird that flies in the hold is not part of it, but located in it. If something is part of something else, it is located in it (axiom bao-1), but again, not the other way round.

That there may be different sorts of parthood is not covered in BFO. The axioms apply to mereological parthood. Perhaps in the future, this will be fine-tuned.

[Werner Ceusters]

For BFO, there is no such formal dependence yet defined. Perhaps in the future, and then perhaps sites and boundaries might become dependent-continuants if dependence of that sort would be true ontological dependence, and not closer like human beings 'depend' on oxygen. It might then fall under one of the existent dependent entities or granted its own category. But it will not done without very serious scrutiny and looking at all consequences, borderline cases, etc. It is better to stay a bit too general, than making mistakes by being too specific. If you need terms like you propose for your work, create them. That is what application ontologies are for. If you want to stay compatible with BFO, ensure that you don't violate any of the current axioms. Or follow the OBO approach: remove axioms. Without axioms, no mistakes and everybody can contribute whatever he wants.

[Alan Ruttenberg]

There's a citation in the BFO 2.0 reference

Barry Smith, “Characteristica Universalis”, in K. Mulligan (ed.), Language, Truth and Ontology, Dordrecht/Boston/London: Kluwer, 1992, 48–77.

I probably read this a long time ago and don't remember any details. If you read it, do report back. I didn't use it as a source for the axiomatization, but maybe there's work that can be done based on it. It says:

x is boundary dependent on y = df (1) x is a proper part of y, and (2) x is necessarily such that either y exists or there exists some part of y properly including x, and (3) each individual part of x satisfies (2).

On first glance this looks like a kind of specific dependence, because of the dependence of x on y existing. There's an "or" but the or references y, so that implies existence. On the other hand, this can't be the case in the current axiomatization because it would imply that the boundary is a specific-dependant continuant, but that's disjoint from boundary, under the assumption that the domain of "is boundary dependent on" is boundary. So let's assume that it is not the case that x or any part of it is necessarily such that y exists. With that we avoid the class with specifically dependent continuant.

"Includes" isn't defined in the paper (it's mentioned in only one other place), but it looks to me like it means  has-part or has-proper-part. Choose the stronger one and substitute:

x is boundary dependent on y = df (1) x is a proper part of y, and (2) there exists some part of y that x is proper part of and (3) each individual part of x satisfies (2).

This would seem to be satisfied if x is any proper part of y, since part of is transitive, and you can choose the part of y referenced in (2) to be y.

Maybe if you read more of the paper you can distill to something that can be added to BFO.

Or maybe I've misinterpreted.

Alan

--
You received this message because you are subscribed to the Google Groups "BFO Discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to bfo-discuss...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/bfo-discuss/fc28ed26-5209-417d-b820-d9ebae700309n%40googlegroups.com.

[Alan Ruttenberg]

On Mon, Jan 23, 2023 at 4:28 PM Thomas Beale <wolan...@gmail.com> wrote:

Initial reaction...

The definition of Site is such that it contains no matter (following the axioms). Since real cavities, holes in cheese, space in a ship's hold etc all contain matter (and may do so in complicated ways, e.g. a sparrow flying in a ship's hold), a Site can only be understood as a geometric specification that has no real-world counterpart, i.e. a 3-D volume (the enclosing 2-D bounding surface would be just as good). It is therefore purely informational and can be represented somewhere, but it does not physically exist. But since that description is also no different from BFO:2-D continuant fiat boundary (other than the quibble about 2-D or 3-D), it can't be that.

Sites don't have material entities as parts. That's different from containment. Containment in a site corresponds to located-in a site, which is certainly allowed. Subsequent arguments that rest on this example are flawed. Werner provided the elucidation of site, which doesn't use the word "contains".

Alan

 

[Alan Ruttenberg]

Long and interesting conversation. I'll add comments here and there, in separate responses. (oops wrote one comment in this vein before this message - didn't mean to be abrupt)

"Without an underlying Material Entity or Immaterial Entity (but not, I think, Spatial region), a Continuant fiat boundary cannot exist - it is a boundary that separates some thing from some other thing."

Not quite. The north pole doesn't necessarily separate something from another thing and it's explicitly given as an example of a fiat point, which is continuant fiat boundary. It could separate two parts of another fiat boundary but the definition doesn't require this. I was bothered by this as well very recently, but the examples set a precedent.

With a formal ontology you need to read the definition rather than rely on your intuition. The definition in ISO says:

A.1.2.24 continuant fiat boundary
ELUCIDATION: b is a continuant fiat boundary means: b is an immaterial entity that is of zero, one or two dimensions, which is such that there is no time t when b has a spatial region as continuant part at t, and whose location is determined in relation to some material entity

It's reasonable to argue that the label is poorly chosen, or that the definition admits things that no one would even consider a boundary. But to use the BFO term, or a term in any well-defined ontology, you have to go by the definitions.It doesn't say anything about separating things. It's also reasonable to subclass the term and provide alternative labels. So you could have a subclass of BFO boundary with the additional condition that it separates two things, and you could choose an alternative label property. Using that label property you could call BFO's boundary something else, and label your subclass boundary, and only present the alternative labels in your application. North pole wouldn't be an example of your boundary class.

Alan

.

[Alan Ruttenberg]

   Tom writes:

Material entity

        Structure - substantial entity whose physical structure independently persists in time (approximately solid state); may contain Material entity/ies; == BFO:Material entity

        Space - cavity created by a Structure(s) (may contain gas, plasma, radiation, nothing) - causally dependent on containing Structure; == BFO: Site

    Immaterial entity

        Materially dependant locus - non-substantial entity defined in relation to a Material entity host, of 1 dim less than the host; == BFO: Continuant fiat boundary

            Material point - point whose location is relative to some part of a Material entity; == BFO: 0-D Continuant fiat boundary

            Material boundary - linear (1D)  on a surface of some part or whole Material entity; == BFO: 1-D Continuant fiat boundary

            Material surface - bounding surface (2D) of some part or whole 3D Material entity; == BFO: 2-D Continuant fiat boundary

        Materially independent spatial locus - non-substantial entity defined with respect to a frame-of-reference, which may be located relative to a Material entity; == BFO:Spatial region

            Spatial point - linear boundary on a surface of some part or whole Material entity; == BFO: 0-D Spatial region

            Spatial boundary - linear boundary on a surface of some part or whole Material entity; == BFO: 1-D Spatial region

            Spatial surface - bounding surface of some part or whole Material entity; == BFO: 2-D Spatial region

            Spatial volume - bounding surface of some part or whole Material entity; == BFO: 3-D Spatial region

I am interested to know what substantive BFO precepts are broken by this reasoning (maybe there are some...)

 

-

What I can do is first evaluate the self-consistency of what you've laid out and then, thanks to the axiomatization, try to see whether what you've written is consistent with those axioms. Some of it can't be evaluated. You have defined new classes "Material entity", "Immaterial entity" that don't have any definitions. In the definitions you have given, which you've associated with BFO classes, you've used relations that aren't defined: "contain" "causally dependent on", "created by", "host". Because they are undefined they don't break anything. Say some more about the relations and we can check.

You haven't said that the classes you've named have mutually disjoint siblings. Can we assume this was intended? Seems to me that given your definitions there can be entities that are both Materially independent spatial locus and Materially dependant locus. Material boundary looks very much like Spatial boundary. How would you distinguish one from the other, given the definitions. The term "substantial entity" is used in important places and isn't defined.

Some of the definitions must have copy-paste bugs. Spatial surface and Spatial volume have the same definition. Spatial point and spatial boundary have the same definition.

In the definition of structure, which you equate with material entity, you suggest (approximately solid state), which is kind of fuzzy but suggests that gases and plasma aren't in the class, whereas they definitely are in bfo:material entity. "may contain", whatever the definition of contains is, just says material entity is in the domain of the relation contains. Space, which you've equated to bfo:site, would seem to rule out "air traffic control region", which is given as an example of site. If I'm correct in these assessments then there are BFO models that would be inconsistent in your framework.

Here's what I think. Don't equate your classes with BFO classes unless they mean exactly the same thing. If you want to retain a relationship to BFO, define new classes, that are, as appropriate, subclasses or superclasses of existing BFO classes. If you want to reuse labels that BFO is using to mean different things, define a new label property, create alternatives to the BFO labels and in your application display using the alternative property. Spend some time defining the new relations you have and crafting statements that involve them that can be turned into formulas, so they can be added to the BFO ones and then checked for consistency.

Bottom line is that since BFO has been formalized, your use of the terms need to be consistent with the formalization. But it's perfectly fine to define new classes relations and relate them to BFO classes.

At the same time, try to find bugs of either omission or entailments in BFO that don't make sense. I'll try to fix those. To the extent that you want to turn text into FOL formulas I can try to help.

Alan

[Thomas Beale]

Alan,

thanks for taking the time to respond (as for Werner). I'll read carefully your responses and respond in time.

A meta-point... I have no argument on the importance of the axioms. It appears that you / WC (and I assume others) are often in the situation of having to convince people that they are the real definition. I have no problem with axioms, I can read CL in my sleep and have built meta-models and interpreters for languages of which CL would be a small subset. No arguments from me. What matters to me is what the axioms mean, for example, what idea of contains, part-of etc is meant. Sometimes there are sufficient axioms to work that out, other times - e.g. undifferentiated 'part-of' etc, there are not.

I do think it is reasonable to be able to discuss what you might consider to be old issues. Otherwise how do things progress? I completely accept that any serious attempt to improve something like BFO (or indeed create a working ontology of any kind) has to be fully axiomatised, and if I were to seriously propose something, that is what I would do, stealing heavily from your work.

One thing I would like to know: which document is considered to be the latest / best version of the Basic Formal Ontology 2.0 Specification and Users Guide and where to find it? Everything else of interest appears to be at https://github.com/bfo-ontology/bfo-2020 .

- thomas

[Werner Ceusters]

What 'parthood' exactly means, is also a question that bothered me. Many papers have been written, in different domains, about possible variations of parthood, etc (is flower a part of bread, etc). Since, the axiomatization of BFO, it bothers me less. What matters is: what happens if you declare two things part of each other in some application, like in an EHR, that uses BFO for some quality control on input?

BFO has no problem with the following scenario presented in Referent Tracking style used as input for a reasoner. The first zero in each sentence means: it is the case that. -1 would have meant, if used: it is not the case that. The rest is similar to the expressions in the CLIF files. The arguments are however not variables. Whenever "tweetie", "t", ... is used in a sentence in particular position, it denotes the very same particular, thus works as an identifier. However, the same particular might be denoted by distinct identifiers.

t([0,instance-of, tweetie, object, t]).
t([0,instance-of, the-mayflower, object, t]).
t([0,instance-of, the-mayflower-hold, site, t]).
t([0,proper-continuant-part-of, the-mayflower-hold, the-mayflower, t]).
t([0,instance-of, tweeties-flying, process, t]).
t([0,participates-in, tweetie, tweeties-flying, t]).
t([0,occurs-in, tweeties-flying, the-mayflower-hold]).
t([0,continuant-part-of, tweetie, the-mayflower, t]).

The reasoner does not complain about the input. Is tweetie part of the mayflower? Well, it was decided to be the case at the given time t by who provided the data, and BFO is fine with it (like some consider pets to be part of a household). What might have been the motivation of the source? Well, tweetie might serve as a carbon monoxide detector. Nobody would object I guess that a factory-made detector would be considered part of the ship at the time it is in the ship. Who disagrees with that, might use located-in, rather than parthood. 

There are on the basis of the axiomatization over 4200 other positive and negative facts generated, plus 835 alternatives about which the reasoner cannot decide, which can in general either happen because of the semi-decidability of FOL, or, in case of this reasoner, which is work in progress, for roughly 99% of alternatives) because it transforms axioms to Kowalski-rules using skolemization what changes semantics. If one of the following alternatives could be proven, either manually, or by another reasoner, the theorem could be added as input for the reasoner. 

The alternatives are:

385    [continuant-part-of,the-mayflower,tweetie,t]
385    [proper-continuant-part-of,tweetie,the-mayflower,t]

Skolemization, however, does preserve satisfiability. If you replace the last sentence in the input above by the following one, it leads to a contradiction.

t([0,continuant-part-of, tweetie, the-mayflower-hold, t]).

Line 1, 3 and 8 are responsible for a contradiction of the following axiom:

  (cl:comment "i is an immaterial entity = Def. i is an independent continuant that has no material entities as parts. [udu-1]"
    (forall (i t)
     (iff (instance-of i immaterial-entity t)
      (and (instance-of i independent-continuant t)
       (not
        (exists (m)
         (and (instance-of m material-entity t)
          (continuant-part-of m i t))))))))

The proof uses also axioms tcd-1 and vbm-1 by means of modus ponens on the generated Kowalski-rules for these axioms

556    ERROR    inconsistency                <--    126     BFO-udu-1     inc    8    33    40
40    TRUE    instance-of    tweetie    material-entity    t    <--    770     BFO-vbm-1     mpp    1        
33    TRUE    instance-of    the-mayflower-hold    immaterial-entity    t    <--    772     BFO-tcd-1     mpp    3        
8    TRUE    continuant-part-of    tweetie    the-mayflower-hold    t    <--    0     unmarked     input            
3    TRUE    instance-of    the-mayflower-hold    site    t    <--    0     unmarked     input            
1    TRUE    instance-of    tweetie    object    t    <--    0     unmarked     input            

Unlike you, I cannot read CLIF in my sleep and a tool like this is handy, even if not perfect, and for sure valuable for teaching better ontology development at MSc-level.

[Thomas Beale]

Just getting back to this after a while. A few questions:

1. Can one obtain the test sets & tools for validating the common logic files? At some point I'd like to play with some modifications, but I'd need to be able to see what breaks.

2. Is the tool you (Werner) used in the post above available for use?

The kinds of things I'd like to experiment with are to answer the following suppositions (which could well be wrong):

A. obtain a satisfying plain-English description of bfo:Site in physical terms. It seems to me that it is a spatial volume defined by a 2-D boundary (i.e. a surface) whose existence at leat partly depends on some material host (Mont Blanc tunnel etc). In the CL axioms it is treated as something that may be a continuant-part-of a bfo:material entity. I continue to think it would be simpler to scrap Site and use 2-D continuant fiat boundary to achieve the same thing (i.e. formalising the idea of an Object being located-in a site but not part of it).

B. I think parthood of immaterial entities (boundaries) within material entities needs to be distinguished from material entity - material entity parthood. Boundaries could almost be argued as being qualities rather than 'parts'. They are like 'edges' of things, i.e. physical discontinuities. For this reason, I think that continuant-fiat-boundary would be better formalised as some Materially-dependent-fiat-boundary (distinguishing it from Spatial region)., with a relation such as has-boundary to connect such boundaries to their Material hosts. Parthood of Materially dependent boundaries would only then be possible between boundary and boundary, not between Material object and boundary. This might be better for proving anything in relation to transitive parthood chains, which currently flows through not just things and their material parts but through all their boundaries as well.

As I say, it might be that I just prove myself wrong, and show that BFO is already optimised! But it would be fun to explore these questions.

- thomas

[Werner Ceusters]

My tools are available for my students in my ontology-related courses and some MS or PhD projects for building full-BFO-based ontologies, i.e. ontologies which take time-indexing seriously. 

The tools are under development, not yet where I want them to be and not well documented yet. I use the students' experiences, mistakes, problems they encountered, etc as input for improvement. Hence I welcome broader use, be it under controlled circumstances.

There are two tools thus far.

1) A parser-converter for axiom collections which are in the exact same format as the BFO2020 axiom files in CLIF. It checks syntactic conformity, performs certain semantic checks, and uses heuristics to identify oddities of various kinds. It generates various output files some of which are files to identify whether there were issues and if so, where. Other output files can serve as input for reasoners or other types of checkers.

2) A reasoner which works with one of the output files as input and checks scenarios such as the tweety story for conformance with the axioms. If it detects non-conformance, there are three possibilities (with combinations thereof): a) the source axioms are not consistent, but this should not be the case when you use only BFO2020 since Alan made sure of that using his tools. Of course, if you add axioms, then the problem might be in the new axioms; b) the scenario-input is inconsistent with the axioms, 3) there is a mistake in either the parser-convertor or the reasoner. It might also be that no conclusion is reached within the time you want it to run, or because the processing exceeds the capacity of the machine it runs on. It also produces various outputs depending on the outcome as partly described in my previous reply.

I am happy to set up a Zoom meeting with you to demonstrate the stuff and discuss how to move forward wrt your goals.

To be clear: I do not have a tool that translates CLIF axioms in 'plain-English' as you write - not too difficult to do but I am not interested in that - nor to translate plain-English statements into CLIF-axioms - very difficult, if not impossible, to do unless you use one or other form of constrained English. I believe John Sowa has developed quite a while ago one such thing to transform constrained English sentences to conceptual graphs. Since CGs are compatible with CLIF, you could then take it up from there.

W

[Thomas Beale]

> To be clear: I do not have a tool that translates CLIF axioms in 'plain-English' as you write - not too difficult to do but I am not interested in that - nor to translate plain-English statements into CLIF-axioms - very difficult, if not impossible, to do unless you use one or other form of constrained English.

That's not my goal with the tools - that's a separate question.

Thanks for the offer - I will get back to you on this.

- thomas