-
Notifications
You must be signed in to change notification settings - Fork 53
Review of material entity branch September 2021
This is a work in progress, based on a subset of the material entity terms. Please feel free to add comments inline, prefaced with your github handle.
There are several major issues.
There are many cases of terms that satisfy definitions but are not subclasses of the term with the definition. If all the subclass relations implied by the definitions are included then there would be a lot of cases of asserted polyhierarchy (APH). In other cases there isn't APH yet, but obvious extensions will require APH.
There are many opportunities to add axioms that are constructed with other terms already in the ontology. A design principle to consider: Any axiom that can be constructed solely from existing terms in the ontology and is true of some term should be asserted.
Disjoints are not asserted anywhere. Disjoints are very informative. For example Armed Force (as defined here) is disjoint with Civil Organization. Service Provider is /not/ disjoint with incorporated corporation.
In many cases defined classes can and should be used as a way to avoid asserted polyhierarhcy.
There are a number of areas, such as bodily component, where it is unclear why a specific small collection of subclasses is included.
In some cases terms can, and IMO should, be imported from existing OBO ontologies. An example is bodily component. Depending on the exact sense wanted a single term can be used, or a defined class can be made. Relevant terms might be a species term from NCBI_Taxon, an anatomical entity term from Uberon, or a defined class based on extended organism in OGMS.
Another opportunity is the overlap between the information branch and IAO. I expect there are more.
Textual definition "A Material Entity that is capable of performing Intentional Acts."
Axiom: "(Animal or Organization) and ('agent in' some process)".
The textual definition suggests a disposition (capable) but the axiom is defined in terms of a process. That would mean that entities are only agents when they are agent in an actual process, which I don't believe to be the intended sense.
In a refactoring I might define:
- Define Agentive process =def process and has agent some material entity.
- Assert the domain of agent in to be Animal or Organism
- Define Agentive disposition = disposition realized only in some agentive process.
- Define Agent as (realized as some Agentive disposition)
Person is not a subclass of Agent but the common sense of the word would seem to satisfy the definition of agent. Unless you mean to exclude incapacitated people. I would recommend direct use of OBO homo sapiens. This term would be imported via MIREOT, but without many or any of the superclassess.
Can be defined in terms of a role realized in recreation or similar, maybe. Generally, defining material entity terms in terms of realizable entities helps avoid asserted polyhierarchy. Wouldn't a park be a site?
Why this selection of body parts? It seems arbitrary. If there's a need to include specific body parts then I expect that would be in a domain ontology, perhaps a slim of Uberon.
Also, do you really mean to use Agent in the definition? Any agent? Vertebrate agents? Animal agents?
- There are sites that would typically be considered environmental features.
- The term is Defined in terms of environment, which is not defined.
- Does the inside of a space station count as an environment? The vacuum surrounding it?
- Is it only planetary features you mean, in which case only earth or any planet?
It's been my experience that feature terms are associated with roles. I expect that feature as defined here implicitly includes some granularity. Is a grain of sand an environment feature, or just a sandy area?
Maybe there needs to be a label change. Or maybe it's a role of something that could be depicted on a map. That's my best guess of what the term connotes.
This kind of distinction is always hard to make. It's been noted in the in the issue 90 debate of natural vs intentional processes.
Better to make it and subclasses defined classes to avoid asserted polyhierarhcy.
The definition seems too broad "An Environmental Feature that is related to or is the result of the influence of human beings on the environment."
The problem is "related to". A person might have a relationship with a valley [site] e.g. located in. That doesn't make a valley an anthropogenic feature.
What about a man-made island? Anthropogenic or geographical feature?
One strategy would be to define geographic feature in terms of being a macroscopic part of a planet + atmosphere, without reference to how it came to be. This would be necessary but not sufficient condition. Then make a defined class of geographic features that are output of an intentional process. This would make some geographical features also anthropogenic features, which addresses the issue of man-made islands. It will also help avoid asserted polyhierachy, as I expect there's an overlap between artifact and anthropogenic feature.
If appropriate, there could also be the condition that it be the sort of thing that would appear on a map, i.e. have a role.
Define it as output of a subclass of planned process, construction. Here is a case where the distinction between has output and has specified output (from OBO). Garbage is also typically output of construction, but it wouldn't be a specified output. Alternatively the constructed part could be related to an Objective.
Is a shoreline a hydrographic, a physiographic feature, or both. I think in any case, should the foundry come to be, environental feature would be delegated to the NGA.
Both Radio Communication Instrument and Telecommunication Endpoint Seems to be a subclass of telecommunication instrument, but isn't asserted as such. Isn't a radio transceiver a telecommunication instrument, going by the definitions?
Telecommunication endpoint: A Telecommunication Network Node that connects a Telecommunication Terminal to a Telecommunication Network.
Telecommunication Network Node: A Telecommunication Infrastructure Artifact that consists of a connection point, redistribution point, or endpoint that is designed to be part of a Telecommunication Network and can be either active or passive.
Telecommunication Infrastructure Artifact: An Infrastructure [artifact] that is designed to support the use of Telecommunication Instruments to communicate.
Infrastructure Artifact: An Artifact that bears an Infrastructure Role.
Infrastructure Role: A Role that inheres in facilities and systems in virtue of their being necessary for the economy of a geopolitical region to function. (note spelling error in original definition "virture").
Compare
Telecommunication Instrument: A Communication Instrument that is designed for use by some Agent in some Act of Communication where the recipient of that communication is potentially a significant distance away from the Agent.
Communication Instrument: An Artifact that is designed to facilitate communication between at least two entities.
Checking:
- It does seem that an endpoint is designed to facilitate communication between at least two entities. (As an aside, is there anything that is designed to facilitate communication between one entity? If not, then the "between at least two entities" is superfluous.)
- It does seem that an endpoint is an artifact that is designed for use by some Agent in some Act of Communication where the recipient of that communication is potentially a significant distance away from the Agent.
I speculated that maybe the difference is that Communication Instrument can't have an infrastructure role, because it isn't a facility or system. But then, neither would it seem that a Telecommunications Switching Node is a facility or system. But maybe it is, with Node being a term of are that denotes a facility. If that's the case then it needs more definitional support as Node doesn't obviously connote a facility.
I think this area needs a combination of some additional work on the definitions and then have the result untangled by using functions and roles to define the terms.
Based on the definition, is a telephone a subclass of Radio Transceiver? Maybe not if the latter can #only# have the received and transmitter. Is a mobile phone a telephone? I ask because these days they also have other functions. If left as is there will be asserted polyhierarchy. To avoid this I think Communication and all its descendants should be defined classes, defined in terms of functions and possibly roles.
Aren't some the other siblings of Radio communication instrument possibly high, very high, or ultra high frequency communication instruments? I think this should be factored make the the X frequency communication instruments defined classes. As it is it will force MI. E.g. depending on what frequencies supported, a ham radio would be both a radio transceiver and a e.g. VHF communication instrument.
As defined, an email box is a material entity. I don't think it is. It is an information content entity, the sort of thing that can be copied from IBE to IBE. Representing these terms as defined classes would allow them to be placed under ICE but still available in queries for communication instruments.
An odd label for intercom (given as definition source). The word Intercom derives from Intercommunication system, but it has it's own sense now. This is a label comment, so I give it less weight.
I think this might better be communicated using a label like local communication system
Interphone more commonly would be called an intercom. But labels, eh? By the definitions, Interphone sounds like it's a kind of telephone.
I'm not sure why public address system wouldn't be a subclass of intercommunication system, given the definitions. Is it because one side isn't a point?
Should be defined in terms of a function to contain. The hint is that "is designed to" generally implies a function.
There is a general issue around Facility and installations that has been raised by David Lutz. I won't repeat it.
All can be defined classes. Facilities that have a function realized as a process that has (specified) output X Aircraft Manufacturing Facility = Facility and realized in a process that has output Aircraft.
No axiom connecting it to government.
I have raised issues with IBE being used where ICE would be more appropriate or fix something that is incorrect, e.g. #94
Some examples:
An Information Bearing Artifact that consist of machine-readable symbols.
Symbols would typically be considered parts of information quality entities unless the sense of "symbol" is something like a paper cutout of e.g. pi. A bar code label would be an IBE. The bar code is an ICE.
Is a picture on a computer monitor an image? What is the instance of Information bearing artifact in that case? Wouldn't it be something else, like the monitor?
Many if not most journal articles are no longer distributed in material form.
Vast majority are purely digital
What are examples of instances? Are they material?
Second comment on issue #94
Several problems have arisen with the proposal that organizations are object aggregates. See BFO issue https://github.com/BFO-ontology/BFO-2020/issues/19 and in particular the resolution retreating from this ideaa https://github.com/BFO-ontology/BFO-2020/issues/19#issuecomment-910391740 A simple solution would move it up to material entity and not use member part of as that would make it an object aggregate.
Axiom: 'continuant part of at some time' some Government. Should it be "all times"?