🚀 go-pugleaf

Lee Sau Dan ~{@nJX6X~} schrieb in Nachricht
<7f3e0spise.fsf@faith.csis.hku.hk>...
>>>>>> "Stefan" == Stefan MUEHLBACHER <SMuehlba@frequentis.com> writes:
>
> Thankyou for your hints concerning http://www.lojban.org/.

>
>    Stefan> So why must the universial language be WRITTEN? I think it
>    Stefan> is much better to invent a universial knowledge or
>    Stefan> language representation. This representation need not be
>    Stefan> understood directly by anybody, it should be an abstract
>    Stefan> representation that can be best handled by computers.
>
>Without anyone understanding it,  how would you expect the programmers
>to implement and *debug* the programs?
>

>
>    Stefan> It should also provide a management for its symbols to keep
>    Stefan> them unambigous without predefining all of them.
>
>Who manages them?  A machine, or a human being?  Does the manager need
>to understand the representations directly?


Maybe I have to explain further considerations about that before I can
answer your questions:

The following is just an idea how global symbols could evolve:

One of the first things needed to represent global knowledge is global
symbols or identifiers.
How is it possible to keep symbols unique when there is not only one
community creating new
symbols but maybe some thousands. Each community uses a set of predefined
standard
symbols and an additional set that they have defined themselves.
Examples maybe companies that define symbols for their specific products or
experts in different
scientific areas or many others.
When two communities merge their symbol set (when merging their
knowledgebases) the result will be:
standard symbols used by both + standard symbols used by A + standard
symbols used by B + symbols specific to A + symbols specific to B.
The symbols specific to A and symbols specific to B have been invented
independently:
So a global symbols naming convention must provide a mechanism to make them
unique.
It is also possible that A invents a symbol different from a symbol of B,
but the meaning of the symbol is the same.
In this case a human finding that out can declare the two symbols identical.
That means: in the merged knowledgebase there is only
one abstract representation for two global symbols,
but now I make a stop.

A simple example of global symbols might be:

global_language
global_language_english
global_language_german
global_country_usa
global_country_france
global_country_germany
global_relation_isa
localA_product_1
localA_product_2
localB_businessunit_1
localB_businessunit_2

In this case a global symbol consists of a sequence of contexts (global,
language) where the last entry (english) is unique within the resulting
context .

When using global symbols it is possible to define relational facts like
this:

global_relation_isa(global_language, global_language_english)

Well, one can understand it, but readability is not very good.

It is not very hard to provide a simple map:

global_language_english: english
global_language: language
global_relation_isa: is_a

Using this map the relation results to:
---> is_a(language,english)

A different symbol order can further improve readability:
---> english is_a language


I can also provide a different map:

global_language_english: Englisch
global_language: Sprache
global_relation_isa: ist_eine

---> ist_eine(Sprache,Englisch)
 or
---> English ist_eine Sprache


------------------------------------------------

Now I can answer your questions:

>Who manages them?  A machine, or a human being?

First of all there should be a naming convention. Then somebody has to keep
the first context unique: This task is similar to URL management
in the internet. New symbols can be invented by a human being or by a
machine that already knows the context.
(The resource description framework is based on a similar idea
http://www.w3.org/TR/REC-rdf-syntax/).

>Does the manager need
to understand the representations directly?

It is very convenient when he does. But when providing a simple map as in
the examples above then the global representation can be hidden.

>Without anyone understanding it,  how would you expect the programmers
>to implement and *debug* the programs?

I am currently debugging a program that uses symbols like that. Now you know
that the symbols are understandable, but not very well.
Providing a map will make it much more readable. Providing another map with
symbols from another natural language will make the
knowledge translatable.
It is clear that the result is not yet a natural language, but it is already
understandable without big learning effort,
because the symbols are already known from a natural language.

>    Stefan> The advantage of this would be: nobody has to learn a new
>    Stefan> language, the translation into a natural language is just
>    Stefan> a matter of symbol mapping and syntax.
>
>Nobody?  Then  who tells the  computers how to translate  the internal
>representations to a natural language?
>


You are right. There must be at least some people that know the internal
representations.
But it is not necessary for the big community to learn them.

>
>
>    Stefan> Is there another reason why your universial language
>    Stefan> should be 'written'?
>
>Any  machine representation  can be  "written" down  on paper  using a
>well-defined format.  On  the other hand, any written  langauge can be
>encoded and  hence represented in a computer.   So, they're equivalent
>in terms of expressiveness.  I  can't see any point in arguing whether
>this universal language should be "written".
>

You are right again. Any representation can be written down. What I meant
here
is that a global representation will be better understood when converted in
the
sense of the maps in the examples above. Therefore its "written" form need
not be used as
main communication medium.

>
>--
>Lee Sau Dan                     $(0,X)wAV(B(Big5)
~{@nJX6X~}(HZ)
>.--------------------------------------------------------------------------
--.
>| http://www.cs.hku.hk/~sdlee                        e-mail:
sdlee@cs.hku.hk |
>`--------------------------------------------------------------------------
--'