Response to Bob Young

From NCOR Wiki
Jump to navigationJump to search

Bob has indeed pointed to two major differences between DOLCE and BFO. Here are the terms in slides 2 and 3 which are of interest for an understanding of these differences:

Slide 2 lid tub

  • 1 million tubs*

Slide 3 hopper silo granule sheet roll (not 'role') mould tool grinding machine waste % of regrind packaging

All but the starred terms refer to objects, in BFO terms, that is to say: to material entities which have a causal unity. '1 million [tubs]' refers to a BFO:object aggregate. I will deal with the object terms using 'lid' as an example. Suppose we have a specific lid, x, that is made of plastic. Then we write:

x made_of plastic =: for all y ((y part_of x & molecule(y)) --> organic_polymer(y))

Molecules, too, are BFO objects. Thus the lid is an object which is a mereological sum of smaller objects, namely molecules. (For simplicity I leave aside impurities and so forth.)

'is' here means 'is identical to'.

DOLCE, I believe, would see the lid as constituted by this mereological sum. So two entities, not one.

Now to:

waste packaging

These are mass nouns (which means you can pluralize them -- you can't say '4 wastes' or '6 packagings'). When I first started working in the medical domain, the ontologies available were full of mass nouns such as these. I noticed over and over again that such terms were associated with errors. Consider for example:

sugar -- meaning the portion of sugar in my tea

sugar -- meaning arbitrary portions of matter made of (one or other) sweet, short-chain, soluble carbohydrate molecules (as in "I always put sugar in my tea")

and

sugar -- meaning the portion of sugar in the universe (as in: "sugar is facing a global supply shortage")

A human can distinguish between these different meanings of the single term 'sugar' and cope in the normal flow of conversation when the word s used without further specification. Once 'sugar' is introduced as a term in an ontology, however, then there will be multiple different contexts where it is applied without humans there to do the needed cross-checking, and errors will very quickly arise.

I worked with various ontologists (including the founders of the FMA Ontology) and, by pointing to examples of errors, persuaded them to banish the use of mass nouns from their ontologies and replace them always by non-ambiguous count noun expressions such as:

this portion of sugar

some portion of sugar

the maximal portion of sugar

and so on.

With this step, the errors connected with mass nouns went away without any drop in expressivity of the ontologies.

If we make a similar proposal for IOF, then we will have to replace 'waste' and 'packaging' with terms like:

portion of waste package aggregate of packages

all of which can be dealt with easily in BFO (and again, without any two-laptops problem).

A portion of waste, in the case described, for example, would be something like a mereological sum of granules (perhaps incorporating some portion of dust, and so forth, mixed in), where each granule is itself a BFO:object that is a mereological sum of molecules. 'is', again, means: 'is identical to'. Thus no two laptops problem.

As to

% of regrind

Unpacking the mass noun we get:

% of portion of regrind

To formalize this we would need to define:

portion of regrind =: material entity that is specified-output-of grinding process

grinding process =: planned process realizing function of grinding machine

To define '% of portion of regrind' we would need to know whether % is calculated via number of granules, mass, or volume. In any case we will have something like a measurement value for the total regrind and a measurement value for the fraction of this total added to hopper, and the ratio of these two would be used to calculate the percent added.

In none of the above do we need to distinguish a manufactured object and the portion of matter of which the manufactured object is made. The two are at any given time identical. Sometimes we deal with portions of matter which are, for instance, mereological sums of molecules; sometimes we deal with portions of matter which are also BFO:objects (because the molecules have earlier been specified inputs to, say, a sheet extrusion process). In talking about portions of matter, therefore, we need to keep careful track of time. (Before and after extrusion, for example.) But this does not mean that after the extrusion we have a second object. Rather, we have the portion of matter shaped in a new form. To deal with this time-dependence non-trivially in a formal way is tricky. But DOLCE's way of dealing with it formally is tricky, also, and DOLCE ends up with the second laptop. Nowhere, I believe, is the second laptop needed to represent engineering processes. Nowhere do engineers concern themselves with this second laptop. The second laptop is an artifact of a certain philosophical view -- which sees constitution in many places where others see identity.

As to Slide 4, the way BFO treats designs and processes, etc., is documented in the BFO (DRAFT) IOF formalization proposal here. I mention the FMA Ontology because it is also the first ontology to document the idea of a canonical ontology which underlies the use of 'planned' (as in 'planned process') in this formalization. See the discussion of 'canonical' in the attached. This use of 'planned' (and associated expressions such as 'has-specified-output') were also incorporated in the Ontology for Biomedical Investigations, which served in turn as the standpoint for the ontological treatment of functionally graded materials sketched here.

I leave it to DOLCE to state how they would deal with designs and processes (the DOLCE version of 2003 does not seem to contain the resources for an ontological treatment of plans).

I realize now that I should have given Nicola more opportunity to explain the DOLCE view in Oslo, and I would welcome the opportunity to combine your slides and the above with a representation of DOLCE views of these matters into a single document, so everyone can compare the pros and cons of the two alternatives.

Critical comments to all of the above are of course welcome.