MIME-Version: 1.0
From: Gleki Arxokuna <gleki.is.my.name@gmail.com>
Date: Wed, 4 Oct 2017 10:51:47 +0300
Message-ID: <CAO7bV+jQVDsVRcJJQpjnxMFETbUZajvHCPGhet=wakdOkvmVYg@mail.gmail.com>
To: llg-members@lojban.org
X-Spam_score: -1.5
X-Spam_score_int: -14
X-Spam_bar: -
Subject: [Llg-members] Old business. Lojban as a knowledge representation or
 programming language
X-BeenThere: llg-members@lojban.org
X-Mailman-Version: 2.1.21
Precedence: list
List-Id: <llg-members.lojban.org>
List-Unsubscribe: <http://mail.lojban.org/mailman/options/llg-members>,
 <mailto:llg-members-request@lojban.org?subject=unsubscribe>
List-Archive: <http://mail.lojban.org/mailman/private/llg-members/>
List-Post: <mailto:llg-members@lojban.org>
List-Help: <mailto:llg-members-request@lojban.org?subject=help>
List-Subscribe: <http://mail.lojban.org/mailman/listinfo/llg-members>,
 <mailto:llg-members-request@lojban.org?subject=subscribe>
Reply-To: llg-members@lojban.org
Content-Type: multipart/mixed; boundary="===============2390878821074641647=="
Errors-To: llg-members-bounces@lojban.org

--===============2390878821074641647==
Content-Type: multipart/alternative; boundary="94eb2c1957f49e278f055ab3e3f6"

--94eb2c1957f49e278f055ab3e3f6
Content-Type: text/plain; charset="UTF-8"

As for Lojban as a cognolang...

*AtomSpace*

Roman Treutlin continues working on English2Lojban converter but the
machine interlingua he uses (AtomSpace) might be underdeveloped, which was
shown by inability to parse made {ko'a .a ba bo ko'e} constructs so fixes
had to be done to Atomese language itself.

Roman made a chat bot for Lojban together with required modules for
AtomSpace:
https://github.com/rTreutlein/LojbanChatBot
but it used PEG grammar [and not any more la tersmu] to parse text which
obscured cognitive transparency of the project (well, English language is
equally terribly analyzed in this project and AtomSpace is the best
humanity has by far implemented).

*Unambiguity is banal.*

Lojban development looks like it's a programming language.
a. unambiguous syntax will save you!
b. ok, it seems not enough. But type checking will save you!
c. well, not enough.

What's next? Well, obviously property-based testing and dependent types but
the problem here is that Coq/Agda/Idris/Scala's shapeless are by themselves
in their infancy.
Nevertheless, we'll have to deal with that one day.
By the way, the most technically rude polymorphism of terbri places was
proposed by Curtis by having a switch in one terbricmi that would specify
what another terbricmi of the same predicate can be filled with. But in
general Lojban and natural languages are already that way if e.g. mlatu1
and gerku1 are of different types.


*Transpiling into proglangs.*

la ramcinfo made a few sketches in regard to transpiling Lojban into
programming languages:
https://mw.lojban.org/papri/User:Ramcinfo/lojbo_staile_liste
#Inspired_by_programming_languages although this is not even a working app
(as compared e.g. to an older Yoshikun's la lojysamban app)

No bridges to RDF or other knowledge representation languages are in
development.


*PEG?*In parsing Lojban grammar nothing exceeding PEG in power has yet been
implemented despite the existence of lots of facilities already used for
programming languages.

In fact the main goal over the last years was de facto ... "not doing
anything new". PEG was first implemented by la xorxes and Robin improving
some things (and deteriorating other things) and unfortunately nothing has
improved since that time.

*Bridges to other types of languages.*
WolframAlpha is a collection of math function in one place with the ability
to evaluate them.
It's not open source but e.g. Sympy and Sage Math are open source
alternatives to it.

Problem: quantum physics/many areas of mathematics are not speakable.
My explanation: it's a Whorfian effect, English simply doesn't have good
words to describe it.
Easy solution: jus translate names of quantum theory functions into Lojban
with each cmima be le'i terbri specified in its type.
Solution: take WolframAlpha's database (or its open source alternative)
where functions and concepts are all expressed in one single programming
language for the whole database. Write a transpiler of that code into
Lojban and back from Lojban.
Farther solution: start optimization of this transpiling algorithm so that
the language becomes and more usable for humans.
As the result we get unification of the language of science and the
speakable daily language.

I have a feeling that this unification can even have pragmatic effects (no
proofs here).


*Resume*
On general this whole subject of cognolangs and proglangs in respect to
Lojban is in its infancy (just like once humans could only say {coi} in
their dialogues).

--94eb2c1957f49e278f055ab3e3f6
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr"><div>As for Lojban as a cognolang...</div><div><br></div><=
div><b>AtomSpace</b></div><div><br></div><div>Roman Treutlin continues work=
ing on English2Lojban converter but the machine interlingua he uses (AtomSp=
ace) might be underdeveloped, which was shown by inability to parse made {k=
o&#39;a .a ba bo ko&#39;e} constructs so fixes had to be done to Atomese la=
nguage itself.</div><div><br></div><div>Roman made a chat bot for Lojban to=
gether with required modules for AtomSpace:</div><div><a href=3D"https://gi=
thub.com/rTreutlein/LojbanChatBot" target=3D"_blank">https://github.com/rTr=
eutlein/<wbr>LojbanChatBot</a><br></div><div>but it used PEG grammar [and n=
ot any more la tersmu] to parse text which obscured cognitive transparency =
of the project (well, English language is equally terribly analyzed in this=
 project and AtomSpace is the best humanity has by far implemented).</div><=
div><br></div><div><b>Unambiguity is banal.</b></div><div><br></div><div>Lo=
jban development looks like it&#39;s a programming language.=C2=A0</div><di=
v>a. unambiguous syntax will save you!=C2=A0</div><div>b. ok, it seems not =
enough. But type checking will save you!=C2=A0</div><div>c. well, not enoug=
h.</div><div><br></div><div>What&#39;s next? Well, obviously property-based=
 testing and dependent types but the problem here is that Coq/Agda/Idris/Sc=
ala&#39;s shapeless are by themselves in their infancy.</div><div>Neverthel=
ess, we&#39;ll have to deal with that one day.</div><div>By the way, the mo=
st technically rude polymorphism of terbri places was proposed by Curtis by=
 having a switch in one terbricmi that would specify what another terbricmi=
 of the same predicate can be filled with. But in general Lojban and natura=
l languages are already that way if e.g. mlatu1 and gerku1 are of different=
 types.</div><div><br><br></div><div><b>Transpiling into proglangs.</b></di=
v><div><br></div>la ramcinfo made a few sketches in regard to transpiling L=
ojban into programming languages:<div><a href=3D"https://mw.lojban.org/papr=
i/User:Ramcinfo/lojbo_staile_liste#Inspired_by_programming_languages" targe=
t=3D"_blank">https://mw.lojban.org/papri/Us<wbr>er:Ramcinfo/lojbo_staile_li=
ste<wbr>#Inspired_by_programming_langu<wbr>ages</a>=C2=A0although this is n=
ot even a working app (as compared e.g. to an older Yoshikun&#39;s la lojys=
amban app)<br></div><div><br></div><div>No bridges to RDF or other knowledg=
e representation languages are in development.<br></div><div><br></div><div=
><b>PEG?<br></b>In parsing Lojban grammar nothing exceeding PEG in power ha=
s yet been implemented despite the existence of lots of facilities already =
used for programming languages.</div><div><br></div><div>In fact the main g=
oal over the last years was de facto ... &quot;not doing anything new&quot;=
. PEG was first implemented by la xorxes and Robin improving some things (a=
nd deteriorating other things) and unfortunately nothing has improved since=
 that time.</div><div><br></div><div><b>Bridges to other types of languages=
.</b></div><div>WolframAlpha is a collection of math function in one place =
with the ability to evaluate them.<br></div><div><div>It&#39;s not open sou=
rce but e.g. Sympy and Sage Math are open source alternatives to it.</div><=
/div><div><br></div><div>Problem: quantum physics/many areas of mathematics=
 are not speakable.<br></div><div><div>My explanation: it&#39;s a Whorfian =
effect, English simply doesn&#39;t have good words to describe it.</div><di=
v>Easy solution: jus translate names of quantum theory functions into Lojba=
n with each cmima be le&#39;i terbri specified in its type.</div><div>Solut=
ion: take WolframAlpha&#39;s database (or its open source alternative) wher=
e functions and concepts are all expressed in one single programming langua=
ge for the whole database. Write a transpiler of that code into Lojban and =
back from Lojban.</div><div>Farther solution: start optimization of this tr=
anspiling algorithm so that the language becomes and more usable for humans=
.</div><div>As the result we get unification of the language of science and=
 the speakable daily language.=C2=A0</div></div><div><br></div><div>I have =
a feeling that this unification can even have pragmatic effects (no proofs =
here).</div><div><br></div><div><b>Resume<br></b></div><div><div>On general=
 this whole subject of cognolangs and proglangs in respect to Lojban is in =
its infancy (just like once humans could only say {coi} in their dialogues)=
.<br></div></div></div>

--94eb2c1957f49e278f055ab3e3f6--


--===============2390878821074641647==
Content-Type: text/plain; charset="us-ascii"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

_______________________________________________
Llg-members mailing list
Llg-members@lojban.org
http://mail.lojban.org/mailman/listinfo/llg-members

--===============2390878821074641647==--