From: Alan R. <ala...@gm...> - 2007-03-01 05:55:25
|
Thanks for this. Some comments: On Feb 28, 2007, at 6:41 AM, frank gibson wrote: > Final Dish, Cooking > Utensil and the rest are all Roles that particular Continuants play > in this > particular Cooking Occurent. Word of caution: "Role" is taken already in bfo to mean something which seems to differ. Here you use Role to mean effectively subproperties of has_participant. There is a discussion of participation in [1] starting at 290. I'm don't think I totally understand the distinctions between siblings Role, Function, Disposition, Power, etc, but I think that it has some to do with the mode of participation or the type of participants. For example, Roles seem to be associated with only people, but that could either be because of agentive participation or because of being a kind of social animal. Anyways, for the sake of consistency, let's note that BFO uses "relation" for this term, and OWL uses "property" and that we should probably settle on one and use it consistently. > A particular recipe can be written down in a particular cook-book, > or in a > PDF, or on a scrap of paper, and in each case it is the 'same' > recipe (if > followed, would mean doing the same things to the same ingredients and > producing the same dish). What differs in each case is that it has > been > encoded into different media. Media is a role that (among other > things) > physical lumps of matter (a page of a book) can play. My understanding is that BFO would relate the information bits to the media by an inheres relationship. There is also this "generic dependence" that was introduced at the OBI workshop to capture the idea that the particular bits in a digital document exist by virtue of their being some number of physical that capture them e.g. 20 hard disks scattered across the world. In any case, in this branch we are not really interested in this aspect, I think, i.e. the relation of the recipe to the document. Rather we are interested from the recipe onwards to the cooking. > Particular recipes are instances of the universal Recipy. When I > 'follow a > recipe', I am instantiating an 'enactment' relationship between the > particular instructions I am following and the particular activeties I > perform. > > In this sense, 'enactment' of a recipe is like 'running' a computer > program. This is a kind of "realization" in BFO. (see [1] 292). Don't know what you mean by universal recipe here. BTW, I think it is a bit muddy where the boundary of particulars and universals is in the realm of realizable entities. BS and I had a bit of discussion about this at the workshop in relation to some OWL prototyping I was doing for the FACS folks. The thing is, a particular plan can be enacted in any number of processes. You can cook a dish many times using the "same" recipe. The sense of "same" is a little ambiguous, as you could imagine instantiating a universal each time the recipe is followed, or instantiating the recipe from a class of curry recipes once and then cooking using it many times. If you used the recipe as a defining characteristic of a class of processes, it is like the difference between using a nominal and class in the definition. > In OWL, we can introduce property restrictions that constrain the > physical > things or activities that can fill the different roles that would be > performed when enacting a recipe. That is, we could state that a > curry recipe > is any recipe where its enactment will be a process that has a > 'final dish' > role filled by some 'stuff' that is a Curry. I think we are on the same page here. So "Cytoxicity Assay" is "protocol" aka "plan" aka "recipe" universal, which is realized aka enacted as a process which have participants which are cells and whatever is involved in the treatment (called the inputs colloquially) and parts which includes - the a process that quantifies the number of cells (yielding information), - a process of applying a treatment and waiting - a process of again quantifying the number of cells. - a number of data transformations which finally yield a ratio which is the fraction of cells killed. Most likely there is also a "control" version of this process (no treatment/sham treatment) In a dose response there are a number of repetitions of this process. One of our challenges is to make clear to ourselves which terms go where. For example, I think cytoxicity assay probably belongs in protocol (plan), rather than process, whereas measurement of number of live/dead cells belongs in process. There are many reasons why one would measure the number of dead cells. Another challenge is arranging things so as to preserve appropriate single inheritance - either in the Rector sense (easier) or the Smith sense (harder, if not impossible). I'm aiming for the Rector sense -Where the ontology only has asserted single inheritance, but defined classes can lead to multiple inheritance after classification. "terms" that we get from the communities may land up either as asserted subclasses or as defined classes. Speak to you tomorrow. -Alan [1] http://ontology.buffalo.edu/smith/articles/cornucopia.pdf |