ontolog-forum
[Top] [All Lists]

Re: [ontolog-forum] semantic analysis was do not trust quantifiers

To: "[ontolog-forum]" <ontolog-forum@xxxxxxxxxxxxxxxx>
From: Pavithra <pavithra_kenjige@xxxxxxxxx>
Date: Thu, 21 Oct 2010 07:27:57 -0700 (PDT)
Message-id: <233773.78859.qm@xxxxxxxxxxxxxxxxxxxxxxxxxxx>
>A 4D model in effect views time as a fourth spatial dimension.   A temporal-
>spatial object thus has an (in most cases) contiguous unchanging 4D
>volume.  Any "change" in the object (from a 3D perspective) is merely the
>difference between two different temporal 3D slices of the unchanging 4D
>volume.


This particular discussion is very ambiguous!   This particular statement above is not very clear !   Has this concept been formally discussed and documented anywhere?  Are they any example what does one model using this model?   What are you modeling using this concept?
This is exactly why I said, that I am not sure about 4D concepts that are discussed in this group.
 

--- On Thu, 10/21/10, FERENC KOVACS <f.kovacs@xxxxxxxxxxxxxx> wrote:

From: FERENC KOVACS <f.kovacs@xxxxxxxxxxxxxx>
Subject: Re: [ontolog-forum] semantic analysis was do not trust quantifiers
To: ontolog-forum@xxxxxxxxxxxxxxxx
Date: Thursday, October 21, 2010, 1:28 AM

In response to:
(comments starting with +)
I found the below in my outbox.  I apparently did not send it a few weeks
ago.

It might clarify things, so i'm sending it now.
-- doug

On 9/30/20106:11 AM, FERENC KOVACS wrote:
> HI, Doug and John
> This discussion has a moving tartget. Unless you accept mental
operations at work, we never get a steady picture in focus.

The miscommunication here seems to derive from John and me discussing
ontology components and Ferenc discussing Natural Language Processing
components.

A general ontology is designed to represent things (and types of things) in
the "real world" as well as mental constructs.  .
(+FK)
+Is general ontology designed? Are mental constructs not part of the “real
world”
 
NL terminology is designed to describe words and linguistic phenomena
+Very vague definition.

>From an ontological viewpoint, words and phrases are linguistic objects
which denote one or many concepts (whether individuals or categories).
Words have different properties from the things they represent.  Words can
be conjugated, declined, pluralized, have tenses, grammatical gender, and
spellings -- all of which are properties most of the things the words
(and phrases) represent do not have.  Similarly, the linguistic objects
do not have the properties that the things they represent do have.

In English, verbs may model relations, but from an ontological view, they
are not relations, merely mental constructs with a language, (probably
unknown) creation date, and set of rules for use and modification.


+Why limit the concept of relations to Boolean relations? Are Boolen relations
not a mental construct?

> John:
> > If you want a 4-D ontology that is compatible with modern
> > physics, I strongly recommend Whitehead's ontology, which
> > makes process fundamental.  What some philosophers call
> > 'continuants' are merely slow-moving processes that don't
> > change their shape very much from one encounter to another.
> > I also discuss Whitehead, among others, in that article
> > cited below.

> I have not read his ontology yet, but I asume that he has a better
> picture of realiyt (hence ontology created) than the rest.

> Although I do not accept the term
> 4D (or 3D for that matter as they are misnomers to me)

They are not NL terms, so do not make sense in that context.
+I do not think that any concepts, such as dimensions, etc. fall outside NL.

> I agree that intdocuing time in an ontology

Both 3D and 4D models deal with time.   A 3D model views a temporal-spatial
object as totally present at any point in time that it exists.  This object
may (or may not) change location and/or shape and gain or loose parts over
time.
 
+This is a messy subject.
    1. Time is a concept defined in relation with space an motion. Both space and
motion are observable so time is represented in terms of space and motion. Since
time is one directional, it is assumed to be of one dimension represented by a
straight line. Since time is assumed to be infinite, this representation is not
correct, because a traight line has an origo, wvebn if we canot see the other
end and assumed it to be ongoing for good. But this is not the case. The
representation of time is therefore done by a cricle, or a cycle with start and
end points at the same location. Thus time is represented by thr face of a
clock, which is a surface, or “2D” in your terms.
    2. To represent time time as motion in space of a dot along a straight line is
satisfactory, and it is still not one dimension to me. You can look at that line
from sideway, it is a line or from the front where it looks only a dot.
Therefore two lines are enough to illustrate motion and time, if they do not
have the same unit of measurement. Why worry about 3D and 4D in connection with
time?

 
A 4D model in effect views time as a fourth spatial dimension.   A temporal-
spatial object thus has an (in most cases) contiguous unchanging 4D
volume.  Any "change" in the object (from a 3D perspective) is merely the
difference between two different temporal 3D slices of the unchanging 4D
volume.

+Well.. Please, try to see directions instead of dimensions. My question: Are
you inside or outside? Do you call your location one dimension or more? Can you
move in three dimensions? How many directions can you move? (do you notice that
you are moving on a surface, which is supposed to be 2D?)
 
is introducing change or movement which is Relation and

Change and movement *can be represented* by relations in a logical system.
But imho, representing any fact, even a static fact, requires the use of a
relation.

Sure, one can create an object that represents the fact, e.g.
TheSinkingOfTheTitanic, but to define it so that logical inferences can
 be drawn requires the use of relations.
+Logical inferences can be drawn a number of ways, not necedssarily based on the
kind of relations used in AI or formal logic.

> which is therefore Verb.

Since an English sentence requires a verb and a sentence is needed to
state a fact (or set of facts) in English, while in a logical notation,
the sentence is mapped to a set of relations operating on arguments, a
correlation between verbs and relations seems evident.
 
+Glad to hear that

>From an ontological perspective, a verb *can* denote a relation which
describes change.  It can thus represent (from a certain viewpoint) a
category of change.  The verb "to be" can also introduce a static fact.

A verb when used in a particular sentence in a particular context can
refer to a particular change (the limits of which might be fuzzy).
+fine

> I do not need more primitives than object, property and relation to
> build an ontology for any domain.

If you are saying that everything can be modeled as an instance of one of
these, i agree.  However, property and relation are subclasses of object.
One must treat properties and relations as objects when defining their
properties.  For example,


+ I do not regard them as sublasses, but I agree that bothe properties and
relations may be regarded as objects which is the result of metal operations
performed on them. I have detailed the process elsewhere myslf.


  (subRelationOf parent ancestor)
  (arg1Isa length SpatialObject)
  (arg2Isa length Distance)

However, the discussion of primitives for an ontology includes more than
what is the minimal root of a class/instance hierarchy.


+I do not think in terms of class/instance hierarchies.
>  To illustrate my points consider this:

> In my view a relation must be embodied by a verb,

This is a discussion of language representation, not ontology.

I translate this as
1. All representations of relations in English must be modeled by verbs.

However, i disagree with this premise, as outlined below.


+Your translation is incorrect. You have jumped to conclusions. ”Do not trust
quantifiers” Forget about all and some and do not pretend that syllogisms equal
logic and reasoning. Far from it.
 
>  because

> a)      a verb is required to represent a predicate

Don't English adjectives and adverbs always represent predicates?  This is
why they are described as modifiers.


+No, they do not. Modifiers belong to verbs, which is the prerequisite of a
clasue, which is a prerequisite for saying something that makes sense.

The phrase, "the fuzzy orange cat on the mat" includes no verbs but
represents three
predicates, one of which is an inter-object relation:
  (color CAT1 OrangeColor)
  (texture CAT1 FuzzyHaired)
  (onTopOf CAT1 FLOOR_MAT1)


+ All that do not make sense in NL. But of course, you can invent all sorts of
languages as long as there are people or machines that are ready to follow.

> b)      a verb is one of the most generic word classes (content words)

So verbs (as well as nouns, adjectives, and adverbs) should be used when
making statements in English (and other languages).


+I did not say that. You can of course simplify NL statements to a form closer
to mathematical notation or formalization, but that exercise should be based on
semantic analysis the statement in NL. For semantic analysis you need to be
concerned with content words only. (Syntax parsing is another story, nothing to
do with it. Notice that adverb is not a content word.)
> where
> c)      the other two word classes (i.e. nouns and adjectives) sem to
> identify objects in spacetime,

Are you distinguishing predicates from relations here?  An adjective
normally modifies a single noun, which if single can normally be modeled
as a binary predicate.


+Now part of the semantic analysis exercise is reality check (the pragmatics, if
you like). In terms of thinking and reasoning however this is the field of
analysis, which concerns not just the representations, but the what is
represented. So you get “metal” analysis, by preforming abstraction, isolation,
etc. for instance to derive properties and relations from objects, and you have
“invasive” analysis by dissecting an objects into pieces (parts). The difference
is crucial and in the latter case you have a version where you have parts that
can be assembled into an assembly and a final product, or you have parts that
can never be fitted together because of the lack of compatible surfaces
(interfaces). With mental operations you do not really remove a property from an
object, but by mainpulating their verbal representations you can do miracles. J)
But comparative adjectives represent inter-object relations, e.g. "elder"
could be modeled by the relation "olderThan":
  "John and the elder Mary"
is represented by the assertion
  (olderThan MARY1 JOHN1)
+ I cannot see the point in scambling a clear NL statement into a spaghetti
language.
>  whereas
> d) nothing but the verbs can represent relations, especially when it
> comes to change and motion behind which you have non-visible entities.

I have shown that prepositions and adjectives can serve this purpose.
The use of an appositive can also do so:
  "Juan and his twin, Maria"
maps to:
  (twins JUAN1 MARIA1)

> e)      Verbs are not possible to grasp the same way as nominals (via
> their boundaries or definitions)

I'm not sure why verbs can not be grasped via their definitions and
boundaries in "the same way as nominals".


+Because verbs by definition express actions, events, states, etc. which
involves motion, change that cannot be pictured by a drawing or a figure, etc.
They essentially involve time and tense, therefore instead of spatial
representation, you need appropriate temporal representation, which is  major
headache, but can be done.
 
What else is needed for grasping them (i.e., their meaning?)?
Is it conjugation and tense issues?


+Conjugation is not important per se. Time and person representation are
important. Clauses make sense, single words do not. Verbs do not make sense
without further specification either.
Why is differential grasping an argument for "a relation must be embodied
by a verb "?
+To reflect reality, which is one of the main purpose of using languages.

>  and
> f)  Verbs seem to need a subject and an object even when they are used
> in SV (non transitive) syntax.

Are "subject" and "object" used here as linguistic terms?   Or do you mean
that a description in a logical language of the use of a verb in a sentence
requires two arguments, the first labeled "subject" and the second "object"?
+In an ontology the elements and concepts I have been trying to outline here  I
do not need that jargon of predicates and arguments, although I can compare all
those formulatiosn (propositions, sentences, declarative and procedural, etc.)

What would the subject and object be for
  "It's raining"?
+Just as in formal languages you need to provide a suitable recursive notation
on the statements, you can use BNF to describe natural language utterances. It
involves transcription, paraphrasing, transformation, etc.
So in the above example you have one form of many possible forms of saying the
same semantic idea, namely “raindrops are falling on my head” . So you have the
choice to see that as: object (rain) relation (rains) property (raining). And
then you can make a dozen or so sentences with that message “as the reflection
of your experience in specific spacetime as defined by the context and your
utterance”

> g) It is the verb that does not make sense (is not complete) without the
> details of person and tense,

Again, this is a linguistic issue, not an ontological one.  I don't see
that it argues for relations having to be modeled by verbs.
+Not my problem
Do you man that linguistic person and tense are inherent in a spoken or
written verb?  This is a property true to some extent of English, to a
greater extent of many other languages (possibly differentially so in
spoken and written language), but not true at all of Mandarin.
+The you need to chose another content word in Mandarinm, which is entirely
different animal anyway as the signs in Chinese are not words in the sense you
have in English.
What are the details of person and tense of the English verb "hit"?
+By lemmatizing words and putting them in a dictionary you have only doen a
morphological analysis. The process you call disambiguation is only
decontextualization and will never get you anywhere near semantic analysis. Frge
is wrong, as I said earlier that the meaning of the individual terminal symbols
derived that way will break you the menaing of the whole string.

Does a noun make sense without plurality, declination, and gender?
+A clause makes sense. A word does not, not even if it is inflected.
> hence it is the only word class to allow us to
> provide information on the generator/identification of any
> communication/message,

What information does the verb "is" in the above sentence provide
about the communicator?  ... or the verb "provide" in the question
i just asked?


+Are we not sharing reality?  
+You mean It’s raining? For the sake of economy and brevity we do not include in
every sentence what is obvious in a communication situation. Otherwise you get
the knots language (Laing, R.D. (1970) Knots. London: Penguin)

When conjoined with a personal pronoun for a subject the verb provides a
little information on the generator of the communication.   In Japanese,
the honorific used also provides such information.
+I believe you, Doug san

> hence compare two statements and find out who is right or
> wrong (what is tue or false).
>
> Doug:
> >>"On Fri, September 24, 20106:22, FERENC KOVACS said:
>> >>...
>> >> Rich:
>> >>> The crucial idea is to NOT start with classes, sets, or other
>> >>> constructions, which are not truly primitive.

>> >>  The infant perceives objects, and situations
>> >> (relationships among collections of objects and situations)
>> >> differently as she
>> >> learns more and more about these mysterious realities.

> > It takes a lot of learning to get to this stage.

> I agree. But if you draw up the path we have been covering new kids
> can follow that with ease and learning will be quicker.

Huh?  The context was whether infants perceiving objects was primitive.
I first state that perception of objecthood is learned, and then provide an
argument below.

I fail to see that anything we are doing will speed up infants learning the
concept of objecthood in the first year of life.
+I should have been more clear here. You use the findings of child
behaviour/learning concepts, like spatial primitives (See Mandler) in order to
use the lessons in later learning, including the augmentation of intelligence,
the acceleration of learning, etc.

> Doug:
> > The infant initially visually perceives patterns of light.  Light is
> > distinguished by color, brightness, and location in the visual field.
> > Some regions of the brain detect specific patterns such as parallel
> > lines and concentric disks.
>
> >  There is no object detection at this point.
>
> > As the head moves the patterns move in the visual cortex.  Early
> > learning relates the patterns moving in the visual cortex to something
> > external to the head.  Motion of the head (when detected) causing
> > motion of visual patterns suggest a fixed existence of whatever
> > causes the patterns.
>
> > Similar motions of portions of the visual field occur when the head is
> > not moving.  This suggests that whatever causes the patterns in the
> > limited sections of the visual field have some existence distinct from
> > whatever causes the patterns in other parts of the visual field.  This
> > seems to be the origin for the attribution of visual objecthood.  The
> > use of other senses helps ground such objecthood.
>
> > Further learning allows objecthood to be assigned to flexible connected
> > motions and continuous objecthood to be attributed to the source of
> > similarly moving patterns with a non-moving (or otherwise moving) region
> > in between (indicative of a separate object between the observer and the
> > moving object).
>
> > Multiple experiences with the same or similar objects allows for
> > classification of objects by type or individuality based on similarity
> > of properties of the visual experience."

> It is all very well. So what?

This was in response to your suggestion that objects are primitive while
types are constructions.  I am trying to show that objects are likewise
constructions of the mind.

Sameness of an object from one viewing to another is also a construction
since it takes a lot of learning for a baby to understand this.
+ we agree on above here
>> FK Maybe.
>>
>> Notice that I do not use concepts like partition, kind, type, etc.
>> as they are
>> not primitives to me. The primitives are still object, property and
>> relation (associated with the dichotome facets above) and they are
>> generated/produced by mental operations

>>DF:
> > Aren't objecthood and relations generated by mental operations?  Some
> > visual properties would be primitive, but most properties are assigned
> > to objects, and thus have to be generated by mental operations as well,
> > since the objects are.
>
> FK: I need to insert two tables here, my appologies for RTF
>
> LORP-words conversion table * Lorp = Lean Object Relation Property model
>
>
> Operations/
> ORP Recognition/
> Identification Isolation/
> Comparison Abstraction/
> Attribution[1] Materialisation/
> Substantiation/
> Name giving Production/
> Sharing
> Object 1 Girl     "girl"
> Property 1   smile smile "smile"
> Object 2  Smile     Smile
> Object 3/
> Product 1         Girlsmile, smilegirl
> VS (syntax)         Girl smiled
> labels         Smiling girl Girl of smile Smiler, girl
> New information Smile
>
>
> Operations Specification Generalization Formalization Interpreation
Production/
> Sharing
> Object 1 The (girl) A girl/girls
>   "g" "girl" Generic/specific ladder
> Property 1   generic Variable/keyword   Descriptors
> Object 2  (her) smile No further properties "s" Change (from no smile
property)
> Property>object conversion
>
> Object 3         Girlsmile, smilegirl
> VS (syntax)     SP (logic)   The Girl smiled
> labels     g+s as tags or key/searchwords
> (Boolean operators)   Smiling girl
> Girl of smile Smiler, girl
> New information Smile
>
>
> ________________________________
>
> [1]A property/attribute and an object is in possessive case (relation) by
> definition. Not always symmetrical though.
>
>
>
>
>> the proof for which is already available
>> in dictionaries, lexicons, etc.
>> But the problem is that the items in those collections (and in other
>> repertories) are sorted morphologically (lexically, or alphabetically) on
>> forms (names, objects) whereas we want a sort on content (properties)
>> which are not available even as a cross reference!!
>
> In an ontology, a sort on property cross reference would be available.
>
> I know. But the problem is that we want a search in terms of properties
> (vontent) instead of form (entry words) in cases when we want to learn
> about the world and/or the words.
>
>> Why? because the way such objects (concepts included)
>> are generated and thus related is not documented - see the
>> debate on process vs. structure. And behind that debate we have the
>> problem of representing verbs (change, motion)
>
> > Verbs are lexical creations.  They represent change (or stasis),
> including change in position or orientations which is called motion.
>
> > I assume you are discussing the problem of representing change (or
> > stasis) instead of representing lexical components.
>
> Verbs can be created from any noun and they are never capable of
> pinpointing anything but a change, which brings in the issue of speed of
> perception and change, the issue of present , past and future, etc.
+It is a pity that the table is ruined above. But I sent you it again in private
as an attachement, if you want to comment on that.

Isn't "to be" a verb which pinpoints stasis, not change?
+You know it very well, that the only way one can grasp somehting as existing
that he has also experienced something as not existing. The Yin and Yang telss
you a lot about it. The eyes work on perceiving contrast, etc.

I can't understand the point you are trying to make below.  It seems that
you are saying that 1D or 2D diagrams (which may include language) can
not represent change, although (1D) language can.  This does not make
sense to me.
 
+I am sorry about that.
    1. Language as written down on  a piece of paper is NOT 1D. This dimension and
direction business seems really hard to follow. A line may be one dimensional,
but it is better portrayed if you say that it is one directional. Therefore
these words coming “in a line” seem to be one directional (dimensional to you)
too, but it is not true. The words are interspersed by spaces, questionmarks,
full stops, etc. that are delimiters of the obejcts between them. Whenever you
stop, you just “change direction”, you move on to another form (word) which is
in fact a NEWLINEstarted, with start and end boundaries.
    2. In looking at a 2D figure your eyes move about to follow the lines. If the
line is straight, you say it is “one dimension”, if  the line is a doodling, you
say it is what? It is 2D, because you need a surface to put your doodling on.


>> in 2D, in a diagram, which by definition is not fit for the job.

> > I wonder what the definition of diagram is that you refer to.  It seems
> > to me that anything (including text) can be added to a diagram to
> > provide needed information.  Are you stating that change can not be
> > described in language?  Written language is 1D; so should be no more
> > fit for the job than a 2D representation.
>
> Just look at the multitudes of diagrams available. (specific diagram
> types) http://en.wikipedia.org/wiki/Diagram
> Neither 1D nor 2D are fir for representing movement.

Language is 1D, so it must not be fit for this purpose, either.


+ I do not ned to repeat what I said above.

> representing
> something in 2D or 1D is drawing something in a linear fashion.  While in
> 1D you chnage directions at the end of the line, it is different with
> pictures.Chunking in writing is done by spaces and symbols (. ! ?, etc.)

Diagrams also use symbols.
+ yes. I meant to say graphics
-- doug

> with diagrams it is different, whichever format is where
> The analysis or interpretation of mental operations start.
>
> ferenc
>
> -- doug
>
>> My best regards, ferenc
>
>
> =============================================================
> doug foxvog    doug@xxxxxxxxxx   http://ProgressiveAustin.org




=============================================================
doug foxvog    doug@xxxxxxxxxx   http://ProgressiveAustin.org

"I speak as an American to the leaders of my own nation. The great
initiative in this war is ours. The initiative to stop it must be ours."
    - Dr. Martin Luther King Jr.
=============================================================

 
_________________________________________________________________
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
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx

_________________________________________________________________
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
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx


_________________________________________________________________
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
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx    (01)

<Prev in Thread] Current Thread [Next in Thread>