Dear Ed and Pat, Ed, I agree partly on this point: [EB] > The point is that there is no shortage of attempts to solve some of these problems. What we need is an assessment of their effectiveness, Yes, absolutely this is an empirical problem that requires testing of hypotheses. The crime is that no such testing (for a diverse community) has ever been funded at anywhere near the level required. SUMO was on a shoestring and that made it much too narrow, without verification in any application. The part that I don’t agree with is the implication that the “primitives” tactic has been tried . Having an “upper ontology” does not meant that anyone has attempted to identify all of the known semantic primitives. That is, IMHO, the most effective tactic. [MW>] Yes an upper ontology does not equate to a complete set of primitives. An upper ontology can reasonably be expected to capture all the ontological commitments, but not all the primitives. I’m not convinced there even is a finite set of primitives. Regards Matthew West Information Junction Mobile: +44 750 3385279 Skype: dr.matthew.west matthew.west@xxxxxxxxxxxxxxxxxxxxxxxxx http://www.informationjunction.co.uk/ https://www.matthew-west.org.uk/ This email originates from Information Junction Ltd. Registered in England and Wales No. 6632177. Registered office: 8 Ennismore Close, Letchworth Garden City, Hertfordshire, SG6 2SU. The other UO’s may be good starting points, but all need supplementation (even CYC’s “Base KB”) to provide what is needed to logically specify the meaning of all other identifiable concepts. More to the point, none of those have been tested to see if they can actually support definitions of all other ontology elements. Pat Pat, (1) The number of mappings will be many-to-1 or 2: my terms to one or two published vocabularies that apparently have relevance and wide use, whereas the universal common language won’t exist. (2) Agreed. So, for the latter situation, we have at least the BFO, SUMO, DOLCE, OpenCyc, and EPISTLE/IDEAS, as repositories of well-defined ‘basic terms’ that can be used to specify meanings. Do we have examples of their effectiveness in resolving such problems at the interface between two “business” activities? In addition, we have several collections of variously defined basic and less basic terms at schema.org and EDIFACT, which can typically be used for somewhat diverse purposes, by virtue of being intentionally somewhat general. Have they been used to resolve subtle and less subtle differences in the usage of terms at business interfaces? The point is that there is no shortage of attempts to solve some of these problems. What we need is an assessment of their effectiveness, and even better, what properties make them effective? I’m not sure whether precision, for example, is part of the solution or part of the problem. -Ed Hans – Creating a set of mappings has two problems: (1) The number of such mappings will be many-to-many, whereas mapping to a common language will be many to one; (2) Many tests have shown that there will often **not** be one-to-one correspondence between terms, and proper interpretation will require that the components of meaning for each term be dissected to recognize the common and the divergent parts of related terms. For the latter situation, an inventory of more basic terms that can be used to specify meanings is essential. That is the role of the common foundation ontology. Pat Pat, Following the logic of the points you make, it seems to me that the focus for promoting interoperability should then be on creating a network-accessible directory of standard controlled vocabularies and any entity’s mapping of its own domain/application-specific vocabularies to one or more of such standard vocabularies (presumably those used by operational partners that entity is likely to encounter in cyberspace and potentially want to interact with). A possible growth industry would then be on collecting experience with such mappings in an increasingly broad range of application domain context attributes and suggestions for improvements in the standard controlled vocabularies. Changes to the latter would presumably be slower and more consensus driven, such as by industry or academic organizations. Hans Clarifying a point commented on by EdB: [EB]: >> I have spent 30+ years attempting to achieve “databases that interoperate with others” AT ALL, to say nothing of “automatically”. “the kinds of variability that people use for terms in general conversation” is enormous. Just getting people to agree on the use and meaning of the terms in a given technical domain can require considerable effort, and the results will often be just “workable”. Typically, one can get ready agreement on 80% of the concepts, with perhaps a few outlier cases. The problem is the 20% that represent differences in experience and viewpoint, i.e., different “theories”. That 20% seems to enter significantly into every corpus of interchange that is key to the successful interaction of diverse agents. The underlying theories/ontologies are at best skew, and at worse seriously inconsistent. What is required is only that, for the universe of discourse of the interactions proper, the inconsistencies do not affect those interactions. Yes, many (most?) people are unwilling to use any vocabulary other than the one they have become accustomed to, and getting a “controlled vocabulary” that is actually universally used properly is probably hopeless. That is precisely why the tactic of using semantic primitives to provide a universal defining vocabulary, and then ****translating**** from the local argot into the common language appears to me to be the best, perhaps the only method to achieve higher levels of interoperability. This only requires that one member of a local team be familiar with the common language and create the logical specifications of the local terms. Of course, those who don’t give a hoot about whether they are understood outside their local community can refuse to expend even that minimal effort, and content themselves with being only dimly understood by others. The rest of us can enjoy the benefits of accurate interoperability, with little effort beyond what is otherwise necessary. You can’t force others to be clear about what they are saying , but if you want to be understood, **you** can be clear, by using some language that is widely used and well-defined. PatC Pat, You only reemphasize what I said earlier. “fundamental” is relative to a theory. What Rich Cooper is talking about is not. I have spent 30+ years attempting to achieve “databases that interoperate with others” AT ALL, to say nothing of “automatically”. “the kinds of variability that people use for terms in general conversation” is enormous. Just getting people to agree on the use and meaning of the terms in a given technical domain can require considerable effort, and the results will often be just “workable”. Typically, one can get ready agreement on 80% of the concepts, with perhaps a few outlier cases. The problem is the 20% that represent differences in experience and viewpoint, i.e., different “theories”. That 20% seems to enter significantly into every corpus of interchange that is key to the successful interaction of diverse agents. The underlying theories/ontologies are at best skew, and at worse seriously inconsistent. What is required is only that, for the universe of discourse of the interactions proper, the inconsistencies do not affect those interactions. People are successful if they are good at recognizing the consistencies in communication, discarding the irrelevant noise that may be inconsistent, and learning from experience what the relevant but inconsistent elements are and how to deal with them. I don’t really know how that relates to the theory of language, but I am reasonably sure it is not about having a basic vocabulary for circumlocution. At least in my trade, it seems to be primarily a matter of learning “where the other guy is coming from” – what you might expect HIS ontology to be. What we see is diverse patterns of business, patterns of practice, patterns of thought, and we learn to formulate ontologies for the foreign patterns that we commonly encounter. In effect, We solve communication problems by reasoning about which ontology They are using and then what kinds of inferences They will make from certain statements. And We find Their ontology (or rather Our approximation of it) by question and answer, or assertion and retort. -Ed P.S. I have no idea whether the above is on-topic or not... Just one point here. Don’t conflate “fundamental” with “universal”. The fundamental (“primitive”) concepts are those that cannot be described (in an ontology, logically specified) by other concepts without generating cycles. That leaves plenty of room, in any linguistic community or computer application, to generate all the terms one needs with only a subset of the primitives. Those are the examples cited over and over again about snow and time concepts in certain primitive cultures. When one is concerned with the “language game” of creating databases that can automatically interoperate with others, the kinds of variability that people use for terms in general conversation is dramatically reduced. If one has a reasonably small (< 10,000) inventory of primitives, it will not be an onerous task for local database creators to create the logical specifications of their terms using the common inventory. It only takes one member of the group who knows how to use the primitive inventory and is familiar with the local domain. I have not seen any other proposed method that has as good a chance of success for that task (“language game”). [EB] > I am quite happy to send the knights errant out on the quest for the holy grail of universal fundamental concepts; it keeps them from ransacking productive villages. Lack of communication among isolated villages also keeps them from being able to use information and products generated by others. Interoperability is worth spending some time on, maybe even riding a horse around. Pat Simon, Thanks for this. Just reading Pullum was a tonic for the day. (His characterization of Whorf alone was worth the read.) But whether some Arctic language has 4 words for snow or four dozen is not Mark’s point. As Pullum himself says, English has 5 or 6 words for wintry precipitation as well. The question at issue, I thought, was whether there is any single notion that lies at the heart of all of them and is “fundamental”, as opposed to a set of closely related concepts, which are not perceived to be more than that. And ‘closely related’ is at heart a term for ‘free association’. What is related is what we perceive to be related: It can be the 6 blind men and the elephant. In the same vein, BTW, I think Rich Cooper’s contention that perception and action are fundamental concepts in the infant mind is at least half in doubt. The universal presence of sensory systems in even primitive creatures suggests that some ‘perception’/’sensation’ concept is fundamental down to some low level of complex life form, and humans are well above that level. But it is my understanding that infants actually ‘discover’ action by sensory feedback. They have to learn all motor control. In the early stages of that process, muscle activation is just another kind of sensation, and the baby learns that he can cause it. Perhaps a “more fundamental” idea is cause and effect. A further linguistic counterexample to ‘fundamental notions’ appeared in a recent National Geographic article. There is a primitive upper Amazon people that appears to have no sense of time. The language has no concept of tense or ‘past’ or ‘future’, and the people themselves live in a very immediate present and lose all interest in any fact or activity that is not continuing within a few hours. I am quite happy to send the knights errant out on the quest for the holy grail of universal fundamental concepts; it keeps them from ransacking productive villages. -Ed P.S. The version of “Eskimo words for snow” that my linguistics professor of 50 years ago offered was that Swahili has 22 words for what we might ‘walking’, because the nature of the physical motion and the surrounding concerns are importantly different, and he offered the list from some study. Conversely, it is a common practice in languages to extend the meaning of a formerly narrower term rather than to create a new one, as in the French extension of ‘noix’ from ‘walnut’ to generic ‘nut’. It is always about what makes a difference (or doesn’t) to the speaker. I should think ‘fundamental’ is what makes a difference to ALL speakers. <http://ling.ed.ac.uk/~gpullum/EskimoHoax.pdf> <http://www1.icsi.berkeley.edu/wcs/> < http://www.yorku.ca/christo/papers/wfdt8.htm> < http://books.google.com/books/about/Science_as_Social_Knowledge.html?id=M16zQgAACAAJ> < http://www.amazon.com/Who-Knows-Quine-Feminist-Empiricism/dp/0877226474> On Oct 3, 2014 9:16 PM, "Mark H Linehan" <mhl@xxxxxxxxxxx> wrote: Regarding "... perception and action are possibly the most fundamental
objects. Therefore I suggest that the vocabulary of sentences communicating
among the agents would have names for designating perceptions and actions,
as initially present in the infant agent ...."
It is well known that different language groups have varying number of
discrete concepts for things like types of snow or shades of colors.
Similarly, different individuals, and groups of individuals, have varying
capabilities for actions and hence varying vocabularies of action.
Therefore, it seems unlikely that there can be a fundamental ontology of
perception or of action.
This is NOT an argument against the idea that "... perception and action are
... the most fundamental objects." It IS an argument against the idea that
there is some "... vocabulary ... for designating perceptions and actions,
as initially present in the infant agent ...."
Mark H. Linehan
-----Original Message-----
From: ontolog-forum-bounces@xxxxxxxxxxxxxxxx
[mailto:ontolog-forum-bounces@xxxxxxxxxxxxxxxx] On Behalf Of Rich Cooper
Sent: Friday, October 3, 2014 3:21 PM
To: '[ontolog-forum] '
Subject: Re: [ontolog-forum] Ontology vs KR
Dear John,
By "handle" I probably should have said
"designate". I am thinking of the handle (a
pointer) you use in a program to indicate the base location of an object
type. The point is, that in looking for fundamentals among human-like
behaviors, you suggested that perception and action are possibly the most
fundamental objects.
Therefore I suggest that the vocabulary of sentences communicating among the
agents would have names for designating perceptions and actions, as
initially present in the infant agent, prior to learning. Learning will add
new words to the kernel vocabulary, layer by layer.
Present technology is fairly good at detecting perceptions of more objective
physical realities, but not at reading psychosocial scenes. Present
perceiving capabilities are not up to human levels in many areas, beyond
human levels in other, and will remain so dimorphic for the foreseeable
future. But they are there, and can be embodied into any agent you may
choose to build.
Actions, by humans, were beautifully shaped by evolution into smooth,
minimal energy-consuming, coordinated movements of the agents effectors,
with feedback from the agent's sensors. When we evolved to plan and execute
more complex actions, the new actions were built as combinations on top of
the kernel actions.
Therefore the infant Kernel of the agent, prior to learning, should include
a vocabulary of each and every perception, and each and every action, plus a
pool of constants, variables and constraints among them, as imposed by the
agent on the environment, and by the environment on the agent.
Learning, based on interaction with knowledge sources (humans, patents,
databases, social networks,...), would of course introduce more and more new
words. Within the realm of patent databases, if word A is called out in a
claim, only As will do. No Bs can just be freely substituted without
demonstrating that B is a true synonym of A, or is an effective equivalent
to A according to the doctrine of equivalents.
So starting with a vocabulary of objects (as
perceived) and actions (as perceived) in claim sentences, the vocabulary can
grow in layers from the Kernel vocabulary up to nearly anything that is
lexically distinguishable. I call each layer a "context", and the IDEF0
model of that context introduces all the constants, variables and
constraints which connect that context to its partitions and to its
immediate parent context(s).
Is that a fair summary?
-Rich
Sincerely,
Rich Cooper
EnglishLogicKernel.com
Rich AT EnglishLogicKernel DOT com
9 4 9 \ 5 2 5 - 5 7 1 2
-----Original Message-----
From: ontolog-forum-bounces@xxxxxxxxxxxxxxxx
[mailto:ontolog-forum-bounces@xxxxxxxxxxxxxxxx] On Behalf Of John F Sowa
Sent: Thursday, October 02, 2014 10:59 PM
To: ontolog-forum@xxxxxxxxxxxxxxxx
Subject: Re: [ontolog-forum] Ontology vs KR
Rich,
The verb 'handle' is extremely vague (or at least underspecified).
In most cases, it means, approximately, "do something with".
JFS
> Any propositional representation in any
language,natural or artificial,
> is an approximation that is based on some
"interesting position on the
> tradeoff". But there is no limit to the number
and kinds of tradeoffs
> for different purposes. Peirce's "twin gates"
of perception and action
> determine the symbol grounding for any and all
representations.
RC
> Then you seem to believe that perception and
action (i.e., embodied agent
> with such) handle all designation of the
vocabulary used to describe what
> was perceived and what action(s) were performed.
The discussions about symbol grounding ask how words and other symbols
relate to the world, directly or indirectly.
Peirce, Wittgenstein,
and others said that the meaning is based on or derived from the way those
symbols are related to perception and action.
For concrete words like 'dog' or 'jump', the connections are direct.
For abstractions like 'justice', the connections are more complex and
indirect. But to be meaningful, an abstract concept like Justice must have
some implications for the way people perceive situations and act within
them.
John
__________________________________________________
_______________
Message Archives:
http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr:
http://ontolog.cim3.net/mailman/listinfo/ontolog-f
orum/
Unsubscribe:
mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/ Community Wiki:
http://ontolog.cim3.net/wiki/ To join:
http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePa
ge#nid1J
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/ Community Wiki:
http://ontolog.cim3.net/wiki/ To join:
http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J |
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J (01)
|