Dear John Bottoms,
Thanks again for a great set of
questions.
I have delayed responding to this email
because I wanted to cover the ground well, because I have been traveling for
the last week, and because I have been sick with something I picked up in the Philippines a
month ago and can’t seem to shake. But now is time to reply, so please
see below,
-Rich
Sincerely,
Rich Cooper
EnglishLogicKernel.com
Rich AT EnglishLogicKernel DOT com
9 4 9 \ 5 2 5 - 5 7 1 2
Rich,
Hmmm, I don't understand your language. I know that it may be common to the
field but I also find that some well accepted concepts can be pitfalls without
an understanding of what is entailed. I don't know Chomsky's grammar. Which of
his many papers can you refer me to?
Many of
Chomsky’s books can be downloaded from the web. Using a Google search on
“Noam Chomsky books download” brings up a lot of links, including:
http://ebookee.org/Noam-Chomsky-Book-Collection_1071034.html
The
following is headed “18 Books and 12 Essays”:
http://torrentz.eu/be40b89ace8664281293cc33311d7f719f21beba
I don't know what
"highly rational text" is, or how it is measured. It seems to imply
that there are other grades of rational text.
That was
my way of saying I think that Chomsky is a very logical writer and he also
explains his theories of how the economic and political world works.
Rationality is comprised of logic and value system, preferring “good”
properties above “bad” ones, and staying consistent (and hopefully
complete) in the logical sense in the author’s conclusions.
The
reason I like those attributes is that I want the simplest text corpus I can
find (simple in logic, in values, in designations and in sentence signatures)
so that my method can be evaluated on the corpus more clearly. Fuzzy writing,
metaphors, and other linguistic cuteness is destructive to an interpreter which
has been programmed to make the best use of the corpus available.
In opening paragraph). You contrast agents and actions ("The first distinction I want to make is between agents and
activities"). I understand this view as it was one that I
once held myself. I tend to think of these things more grammatically now.
Given
the past history of results which English grammar has produced in practice, I
am not so satisfied. For example, I sometimes use the Link Grammar Parser
(LGP) which lets me calculate all the possible parse trees (within the loaded
grammar) that can be made of a sentence. Its computing time and memory bounds
are too high, IMHO, to make it a practical delivery mechanism for the interactive
use of language. After many years of using BNF context free grammars, I find
it is more productive in MOST use cases to instead keep a very simple grammar
and use a case base to disambiguate the many interpretations.
The problem arises with
the understanding of "actor" or "agent" vs things and
activities. Agents don't always do things, they may not take an action or may
be inhibited from taking an action. They are in effect "nouns" that
do "verb" things.
While
true, agents in my firmament are also acted UPON by activities. For example,
in an activity that trains a customer, I get an input which identifies the
agent (customer), applies a control (training course, web site ..) and produces
as its output a customer in the trained state.
By
limiting the top level to Agents and Activities, I can fit into the IDEF0
process very nicely. For example, even a rock is acted upon by geological
forces. By representing the rock as an activity instead of an object, I can
model the process of weathering, wearing down, and cracking into parts while
still treating the rock as a simpler object in short term activities which are
only concerned with immediate use of the rock in other activities. So this is
a design decision I have made rather than an ontological decision about how to
represent rocks. I want the most general representation that can be further
qualified, constrained and acted upon rather than the most ontologically
sophisticated representation.
This view is confounded by a view of things (nouns) that do things when they
are not supposed to. I first ran into this problem in a linguistic setting;
note that you do not see this in the semantic analysis, but may in a linguistic
analysis. Take for example: a rock (clearly an inert thingy) sitting on a clay
slope. Along comes the rain and it no longer is a thing but becomes an agent
but perhaps not quite an actor. Anyway, it slides down the slope unlodging
others of its kind with a few logs. It seems to act on its own and starts to
look more and more like an avenging angel agent.
Yes, but
rock is not an Agent because it doesn’t “do” anything. It is
only acted upon by other forces through Activities performed with it, on it, or
to it. If I were to say that a rock is an Agent, it would have to have
motivation (self interest), plans, beliefs, and other qualities that agents
have so that it could represented efficiently in a relational format.
My suggestion is to stay
with verbs and nouns for as long as possible. This puts you in the syntactic
arena but that is safe as long as it can be tolerated.
Agreed,
but verbs are associated with Activity definitions and instantiations that can
be more deeply expanded in a relational DB, queried in appropriate ways,
grouped with other like verbs, and qualified in groups or individually. That
is useful in searching for linguistic representations in an AND/OR tree.
So I use
the slightly more abstract term Activity and instantiate Activity definitions
for each verb use case.
I understand that at some
point you may need to move to semantic analysis, but resist until you can't resist
any further.
As I
explained above, I don’t use syntax as a formal analysis tool, but as a
sequence analysis tool on small phrases. The LGP experience I have had leads
me to believe that is more effective than using currently available parsers at
the FIRST stage of semantic analysis. I prefer a case base method which uses
relational DB technology as explained in the ‘923.
I don't have the expertise to read the patent expediently.
My other comments are embedded below.
-John Bottoms
FirstStar Systems
Concord, MA USA
On 5/30/2012 5:22 PM, Rich Cooper
wrote:
John, those are excellent questions for describing the
context of this project. My comments are below,
-Rich
Sincerely,
Rich Cooper
EnglishLogicKernel.com
Rich AT
EnglishLogicKernel DOT com
9 4 9 \ 5 2 5 - 5 7 1 2
Rich,
I find it useful to start the development with questions that define the scope.
What is the problem to be solved.
This formulation is a
system spec for a reference design that practices my patent 7,209,923, attached
to this email. A previous reference design was developed to process
patent claims from the USPTO patent database.
This reference design
(TBD) is intended to demonstrate the ability to analyze highly rational text,
which Chomsky typically writes (though not always). The idea is to manage
a lexicon of verbs (including gerunds and signatures a la Beth Levin) that
relate to activities which individuals perform, as described in the corpus
text.
The first distinction I
want to make is between agents and activities. With the signatures there
are embedded variables that can be bound to the actual objects and phrases that
are not verbs or gerunds according to the initial lexicon. Chomsky’s
language is so precise and consistent (IMHO) that his texts provide a
foundation for simpler accumulation of signatures for verb phrases that might
be useful initial lexicons for specific applications TBD later which also
practice the ‘923 methods.
1. Is this taxonomy looking outward?
I’m not sure what
you mean by “looking outward”; please elaborate so we are talking
the same concepts.
By
"outward" I mean from the self to another agent. The difference
between self-interest and other-agent's-interest is in the missing information
for the other-agent. So the processing is different.
Is there someone's (or
some group's) motives I am trying to understand?
Yes, Chomsky’s
favorite subject relates to political, economic and military actions performed
by politicians, government employees, corporations, NGOs and at least those
other agent classes.
Does
this pertain to the patent?
(As an aside, I believe there are entre's to duTocqueville's "hidden
hand" concept if the duality analysis is pursued carefully. That would be
an interesting study.)
This view would be useful
in duality evaluations.
Please define
“duality evaluation”; I’m not sure what you mean by that
phrase.
Nah,
I made it up. I am referring to the ability to develop a theory for an agent
based on behaviors. Of course, to do it right one has to consider oneself as
part of the situation.
2.
For myself, looking inward; what is my goal and what are the range of behaviors
or processes that are candidates to further my position vis à
vis my goals.
My goal is to have a
reference design that can be adapted to a wide variety of applications, and
which can be attractive to licensees of the patent. The main attraction
is to provide a minimal reference design that can start an application designer
on the path to practicing the invention.
3. What problem is trying to solve. That is: what is the current restrain on
this self-interest. This would lead to a change of analysis of restraints.
By reducing the cost and schedule of a considered
application, the reference design helps potential licensees get a jump on
minimizing the cost and schedule for the application. That economy is in
the licensees self interest.
It
sounds like you are trying to develop a grammar for an application. Is that
right?
Re the self interest of
the corpus and the agents mentioned there, I think the best approach is simply
to build a case base of the behaviors which each agent performs (i.e., which
activities and which object bindings). The case base in this patent is
stored in a database that can be dynamically sized and configured (see the
‘923 specification for an explanation). Finding patterns of
linguistic use within the parsed and stored database is a lot easier than
constructing the software to make and manage that database in the first
place.
It
sounds like you need a meta-grammar application to find and label the patterns.
This is done for Hidden Markov Model systems that are typically hand-tweaked.
Statistical analysis such as in Watson can help, but to be safe you need to
have a human review the results of any machine learning algorithm.
For me, I don't see much of a difference between patterns and algorithms at
some point. There appears to be a mathematical mapping from one to the other.
Watson has contributed by identifying the important subject areas under the
query curve, at least for Jeopardy. Clearly, it doesn't understand the subjects
being discussed, and in that sense it is my opinion that it is a
"statistical gossip machine".
4.
'Course the "cost/benefit" is elemental, including community standing
as a self-interest item (both in terms of success or failure).
Again, the behaviors
(actions) of the agents can be organized into a case base and analyzed to infer
self-interest and coping behaviors from the database of text phrases and IDEF0
interpretations.
I'm
not as fully studied in behavioral economics as I would like to be. My exposure
in this area is via assessment, which is an adolescent field. I would encourage
you to include metrics linked to scoring and behaviors to understand another's
self-interest traits.
5.
History: how has this been solved before, why can't that be done now. Precursor
contexts may be interesting: how did we get here, why does this problem need to
be solved.
Agreed; the Before and
After conditions of each activity should be linked through database indexes to
identify patterns, infer self interest and coping behaviors for each
agent.
Pre-
and -post scoring in statistics is better understood than normative assessment.
The cutting edge currently is in "adaptive polytomous assessment".
These extract multiple traits while selecting against a controlled variable.
These techniques require significant investment until better tools are
developed and there's not a lot of call for them now. Again, the proximal
region hasn't required it just yet.
6.
Future: how is this going to affect the context, who are those in the context
that will be effected
Context is the
represented element of an IDEF0 activity that is organized into the database as
associated ICOMAs and their component object types and decompositions.
;TL,
DR
7.
Time, how long will it take, do the parties have that much time?
The plan, as in 3 above,
is to provide a starting structure for reducing the cost and schedule for
developing future applications. Not every writer is as precise and
accurate as Chomsky, so metaphors, incomplete sentences, ambiguous sentences
and phrases, and other related application elements will show up, but ways to
handle those derived considerations still will have to be done individually,
for each application, on top of the reference design.
My
comments here referred to the length of time an assessment takes. That may not
be relevant in your case if the assessment can be handled strictly with
semantics aside from time considerations.
-John
Bottoms
FirstStar Systems
Concord, MA
Thanks for your
comments; they were very useful. Please keep them coming,
-Rich
On 5/30/2012 1:41 PM, Rich
Cooper wrote:
Ed
Barkmeyer wrote:
This is
a would-be taxonomy without classification criteria, or
consequent
properties, which makes it ontologically useless.
Alternatively,
one can see it as an itinerary of places to be visited on
the way
to formulating some kind of political science ontology.
-Ed
True, it’s only a
starting point, not a finished ontology. The final version should
ultimately be capable of reasoning in its various shades.
I still want to map the
elements cited (and others TBD) into a regular rendering, and I prefer IDEF0
since it is well known by nonontologists. When the ontology is finished
some day, it can be mapped into the IDEF0 structures and interconnects, and
augmented with rules for logic, etc.
This is
a would-be taxonomy without classification criteria, or
consequent
properties, which makes it ontologically useless.
Alternatively,
one can see it as an itinerary of places to be visited on
the way
to formulating some kind of political science ontology.
True. Its only a
starting point.
Your discussions below re
the meaning of self interest is also useful. I’ll think about it
some more, but I think self interest is still a slippery concept to me, and
will take some deeper thought to render properly.
Thanks for the thoughts;
more would be appreciated also,
-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 Ed Barkmeyer
Sent: Wednesday, May 30, 2012 8:52 AM
To: [ontolog-forum]
Subject: Re: [ontolog-forum] Self
Interest Ontology
Ignoring the choice of
political scientist as source, this purports to
be the basis for an
ontology.
What is the basis for this
high-level taxonomy? What are the
distinguishing properties
of the 2nd level classifiers?
And how do those
properties relate to "self-interest"?
All organizations have
self-interests, and their human components
typically share some of
those interests, either out of ideology -- what
the organization does is
"good" -- or out of pragmatism -- I will do
well only if the
organization does well (even though the definition of
"x does well" is
quite different for x=me vs. x=the organization). But
there are also cases in
which the self-interest of the individuals may
be to the detriment of the
organization -- I can use the organization to
do well -- as in the
leveraged buyout game. And a lot of work is
motivated by "pride
in accomplishment" -- personal fulfillment, which
may or may not be
"useful accomplishment" -- organizational value.
Bureaucrats commonly
mistake making a contribution for making a useful
contribution. So, it
is not at all clear to me how position in an
organization affects the
self-interest of either the individual or the
organization.
This is a would-be
taxonomy without classification criteria, or
consequent properties,
which makes it ontologically useless.
Alternatively, one can see
it as an itinerary of places to be visited on
the way to formulating
some kind of political science ontology.
-Ed
Rich
Cooper wrote:
>
> If we limit the Self
Interest Ontology to just the players Chomsky
> mentions (directly or
indirectly), the set of agents could be
> organized thusly:
>
>
>
> -government
>
>
-legislators
>
>
-judiciary
>
>
-cabinet level executives
>
>
-employees
>
> -NGOs
>
> -corporate
>
>
-stockholders
>
>
-directors
>
>
-executive management
>
>
-lobbyists
>
>
-employees
>
> -individuals
>
>
-taxpayers
>
>
-adults
>
>
-minors
>
>
-beneficiaries
>
>
-adults
>
>
-minors
>
>
>
> This gives one view
for the Ontology which identifies the agents that
> participate in the
Chomskyesque materials. Does anyone want to
> suggest additions,
deletions, or changes to the list above?
>
>
>
> Activities of the
Self Interest Ontology might include:
>
>
>
> -government
>
>
-taxation
>
>
-regulation
>
>
-legislation
>
>
-enforcement
>
>
-judgments
>
> -corporate
>
>
-markets
>
>
-monopolies
>
>
-competitors
>
>
-persuasion
>
>
-operations
>
>
-finance
>
>
-capital
>
>
-revenues
>
>
-costs
>
>
-lobbying
>
>
-employment
>
>
-taxation
>
> -Individuals
>
>
>
> One way to develop
materials for filling in the lower levels of the
> ontology might be to
process NLP from Chomsky’s books and articles,
> and news stories,
including daily news articles from individual
> reporters, articles
from corporate news sources (e.g., WSJ, NYT,
LA
> Times, etc). By
identifying the named entities that correspond to the
> agent classes above,
it should be possible to organize news stories to
> deepen the Self
Interest Ontology to include lower level subclasses.
> This is a very
limited first step in identifying the actors and
> activities that play
identifiable roles a la Chomsky’s viewpoint. It
> should also identify
the news sources which are biased in each
> direction for each
class of agents.
>
>
>
> Has anything serious
been left out of the top level for the Self
> Interest
Ontology? Again, suggestions are appreciated,
>
>
>
> -Rich
>
>
>
> Sincerely,
>
> Rich Cooper
>
> EnglishLogicKernel.com
>
> Rich AT
EnglishLogicKernel DOT com
>
> 9 4 9 \ 5 2 5 - 5 7 1
2
>
>
------------------------------------------------------------------------
>
> *From:* ontolog-forum-bounces@xxxxxxxxxxxxxxxx
> [mailto:ontolog-forum-bounces@xxxxxxxxxxxxxxxx]
*On Behalf Of *Rich Cooper
> *Sent:* Wednesday,
May 30, 2012 6:46 AM
> *To:* '[ontolog-forum] '
> *Subject:* Re: [ontolog-forum] Self
Interest Ontology
>
>
>
> Chomsky’s
theories of the corporate-state partnership, and how it
> concentrates power in
the hands of large corporations, are well known,
> especially:
>
>
>
> http://www.youtube.com/watch?v=TieGj2Yi5r8
>
>
>
> But neither Rand nor
Hayek subscribed to corporate-state partnership.
> In both cases, they
value the individual, not the corporation and not
> the state and
certainly not the combination of the two. So I don’t
> think that is the
reason why he is unhappy with both Rand and Hayek.
>
>
>
> From the above video
“elections are always bought”, “President Obama’s
> election was funded
by corporate interests”, and numerous other
> examples indicate his
deep displeasure with the state-corporate binding.
>
>
>
> I would like to see
quotes from Chomsky that specifically describe his
> displeasure with both
Rand
and Hayek rather than trying to predict his
> rationale.
Chomsky is always very deep in his rationale, so I don’t
> think that we can
simply say the corporate-state binding is why he is
> against either Rand
or Hayek.
>
>
>
> Mike Pool quoted this
short paragraph from the article, which is
> somewhat enlightening:
>
>
>
> "Hayek was the
kind of 'libertarian' who was quite tolerant of such
> free societies as
Pinochet's Chile,
one of the most grotesque of the
> National Security
States instituted with US backing or direct
> initiative during the
hideous plague of terror and violence that
> spread over the
hemisphere from the 60s through the 80s. He even sank
> to the level of
arranging a meeting of his Mont Pelerin society there
> during the most vicious
days of the dictatorship. "
>
>
>
> But that critique is
not directed at Hayek’s ideas about economics in
> general, only about
his interpretation of Hayek’s poor showing in the
> political area,
specifically in supporting Pinochet and the US
methods
> of supporting
property owners at the expense of the average citizen.
>
>
>
> His description of
how democratic groups in Haiti
were overturned by
> the US
government and a dictator was reinstalled under US actions, is
> typical Chomsky, and
very clearly in line with his past work. But in
> broad brush strokes
in the article, he paints both Rand and Hayek
> (neither of whom are
known for their political wisdom) as evil without
> considering the kudos
they gave to the individuals.
>
>
>
> -Rich
>
>
>
> Sincerely,
>
> Rich Cooper
>
> EnglishLogicKernel.com
>
> Rich AT
EnglishLogicKernel DOT com
>
> 9 4 9 \ 5 2 5 - 5 7 1
2
>
>
------------------------------------------------------------------------
>
> *From:* ontolog-forum-bounces@xxxxxxxxxxxxxxxx
> [mailto:ontolog-forum-bounces@xxxxxxxxxxxxxxxx]
*On Behalf Of *Chris
> Menzel
> *Sent:* Monday, May
28, 2012 5:30 AM
> *To:* [ontolog-forum]
> *Subject:* Re: [ontolog-forum] Self
Interest Ontology
>
>
>
> On Mon, May 28, 2012
at 2:09 PM, Rich Cooper
> <rich@xxxxxxxxxxxxxxxxxxxxxx
<mailto:rich@xxxxxxxxxxxxxxxxxxxxxx>>
wrote:
>
> Thanks – a very
interesting article. I’m surprised at how vehemently
> Chomsky shrugs her
off as evil. He doesn’t give any explanation in
> the article; do you
have any information about WHY he thinks Rand is
> evil?
>
>
>
> It's obvious if you
read Chomsky's (vast) work on political theory and
> American social and
political history, especially his writings on
> social security,
taxation, corporate welfare, the massive
> redistribution of
wealth from the middle class to the top income
> brackets engineered
by conservative tax policy over the last dozen
> years, etc, all of
which are in vehement opposition to the social
> darwinism that lies
at the heart of Randian economic theories (and
> current GOP economic
politicies).
>
>
>
--
Edward J.
Barkmeyer
Email: edbark@xxxxxxxx
National Institute of
Standards & Technology
Manufacturing Systems
Integration Division
100 Bureau Drive,
Stop
8263
Tel: +1 301-975-3528
Gaithersburg,
MD
20899-8263
Cel: +1 240-672-5800