[lojban] (Technical) Problem area in v3 grammar

17 views
Skip to first unread message

Richard Curnow

unread,
May 19, 2000, 6:39:55 PM5/19/00
to Lojban List
coi rodo

Consider the following (contrived) example :

"Before the month in the springtime, I went to the house"

Possible translations :

pu le masti pe zi le vensa mi klama le zdani

or maybe if we paraphrase a bit

pu le masti ezibo le vensa mi klama le zdani

or perhaps if we paraphrase a bit more ...

pu zi gi le masti gi le vensa mi klama le zdani

Unfortunately not! The pu is 'absorbed' into the tense associated with
the 'gi' construct, so the whole pu..vensa phrase becomes the x1 of klama,
instead of 'mi', which becomes x2.

In my makefile for jbofi'e, I have a target where all the grammar
rules with optional terminators are removed, leaving only those with
the terminators in. This sumti-tcita + tag/gi/gi sumti forethought
connective case turns out to be the only remaining shift-reduce conflict
in the grammar after the pruning : there is no way to terminate the 'pu'
tense before starting the tense that is the connective associated
with the gi/gi pair.

BTW, is anyone taking ideas yet for things to look at after the 5
year freeze is up? What is the methodology for proposing, debating
and selecting potential changes? [I don't expect there will be
many significant ones, I'm just wondering what happens if there are any
at all.] Following a thread a week or so back, the only other topic
I've thought about (besides the problem earlier) is to start using the
CV'VV cmavo space to build abbreviations for frequently occuring strings
of cmavo, e.g. ca'ai for 'ca le nu' and so on.

co'o mi'e ritcyd.


--
----------------------------------------------------------------------
Richard P. Curnow r...@myself.com
Weston-super-Mare
United Kingdom http://www.rrbcurnow.freeuk.com/


------------------------------------------------------------------------
WRITERS WANTED! Themestream allows ALL writers to publish their
articles on the Web, reach thousands of interested readers, and get
paid in cash for their work. Click below:
http://click.egroups.com/1/3840/3/_/17627/_/958776044/
------------------------------------------------------------------------

To unsubscribe, send mail to lojban-un...@onelist.com

Bob LeChevalier (lojbab)

unread,
May 20, 2000, 3:40:37 PM5/20/00
to Lojban List
At 11:39 PM 05/19/2000 +0100, Richard Curnow wrote:
>coi rodo
>
>Consider the following (contrived) example :
>
>"Before the month in the springtime, I went to the house"
>
>Possible translations :
>
>pu le masti pe zi le vensa mi klama le zdani
>
>or maybe if we paraphrase a bit
>
>pu le masti ezibo le vensa mi klama le zdani
>
>or perhaps if we paraphrase a bit more ...
>
>pu zi gi le masti gi le vensa mi klama le zdani
>
>Unfortunately not! The pu is 'absorbed' into the tense associated with
>the 'gi' construct, so the whole pu..vensa phrase becomes the x1 of klama,
>instead of 'mi', which becomes x2.
>
>In my makefile for jbofi'e, I have a target where all the grammar
>rules with optional terminators are removed, leaving only those with
>the terminators in. This sumti-tcita + tag/gi/gi sumti forethought
>connective case turns out to be the only remaining shift-reduce conflict
>in the grammar after the pruning : there is no way to terminate the 'pu'
>tense before starting the tense that is the connective associated
>with the gi/gi pair.

I'll have to think about this. My preliminary guess is that this will be a
limitation on forethought tense constructs (which aren't the most used
things in the grammar anyway). John Cowan will of course have to opine.

>BTW, is anyone taking ideas yet for things to look at after the 5
>year freeze is up? What is the methodology for proposing, debating
>and selecting potential changes?

The concept of the freeze is that we will not even take ideas, propose them
or debate them during the time of the freeze (which is open-ended - it is a
5 year MINIMUM in which the 5 years starts only after the books are all
published, which still has not occurred yet).

We of course can't or won't stop discussions on the subject, but LLG has a
firm position to not consider changes during the baseline. I am not sure
we will even consider a methodology for changes during the baseline.

> [I don't expect there will be
>many significant ones, I'm just wondering what happens if there are any
>at all.]

If there are any at all, we ignore them for now, and think about it later
if they somehow stick around or resurface later. With any luck,
discussions in that later time will be required to be in Lojban, so that
decisions are made internal to the language, but I have given no more
thought to the subject than that.

> Following a thread a week or so back, the only other topic
>I've thought about (besides the problem earlier) is to start using the
>CV'VV cmavo space to build abbreviations for frequently occuring strings
>of cmavo, e.g. ca'ai for 'ca le nu' and so on.

That space is experimental; it is not a baseline violation to experiment.

lojbab


----
lojbab loj...@lojban.org
Bob LeChevalier, President, The Logical Language Group, Inc.
2904 Beau Lane, Fairfax VA 22031-1303 USA 703-385-0273
Artificial language Loglan/Lojban: http://www.lojban.org


------------------------------------------------------------------------
72% off on Name brand Watches!
Come and buy today and get free shipping!
http://click.egroups.com/1/4011/3/_/17627/_/958851526/

Richard Curnow

unread,
Jun 3, 2000, 5:32:22 PM6/3/00
to loj...@egroups.com
(Apologies if this looks a bit cryptic - I sent a private mail to John
yesterday about this topic - the citation is taken from that and lacks
the previous context.)

On Fri, Jun 02, 2000 at 06:31:07PM -0400, John Cowan wrote:
> Richard Curnow wrote:
>
> > term<83> = sumti | (tag | FA #) (sumti | /KU#/ | termset) | termset | NA KU #
> >
> > Yes? No?
>
> Yes.

Having had another day to think about this, I think John's suggested
modification to grammar.300 provides more generality than is actually
required to fix the problem.

I believe the minimal change needed is to redefine term<83> to look like
this :

term<83> = sumti | tag /NUhI/ sumti | tag /KU#/ | FA # (sumti | /KU#/) | termset | NA KU #

i.e. make it optional to insert nu'i between a tag and the following
sumti. This provides the delineation required between the pu and the zi
in my original problem example. Probably lots of other cmavo would be
useable instead of nu'i without problems - I've just picked nu'i to
mirror John's suggestion most closely.

Whether this is easy to do in grammar.300, I don't know. It does have
the advantage of fixing only the buggy case. It could be trivially
implemented in jbofi'e (as could John's proposal).

>
> > I'm interested in this because the parser in my jbofi'e program was
> > written from the bnf.300 file - I've avoided looking at grammar.300 and
> > your lexer + missing terminator insertion code altogether, to try and
> > keep my program completely indepedent.
>
> Okay. In case of conflict, though, grammar.300 is authoritative.

But as I mentioned in my previous mail to John, it's useful to develop a
parser from bnf.300 alone, if only to give an further check of whether
bnf.300 and grammar.300 describe the same grammar.

co'o mi'e ritcyd.

--
----------------------------------------------------------------------
Richard P. Curnow r...@myself.com
Weston-super-Mare
United Kingdom http://www.rrbcurnow.freeuk.com/


------------------------------------------------------------------------
WRITERS WANTED! Themestream allows ALL writers to publish their
articles on the Web, reach thousands of interested readers, and get
paid in cash for their work. Click below:

http://click.egroups.com/1/3840/3/_/17627/_/960068573/

John Cowan

unread,
Jun 2, 2000, 6:31:07 PM6/2/00
to Richard Curnow, loj...@egroups.com
Richard Curnow wrote:

> term<83> = sumti | (tag | FA #) (sumti | /KU#/ | termset) | termset | NA KU #
>
> Yes? No?

Yes.


> I'm interested in this because the parser in my jbofi'e program was
> written from the bnf.300 file - I've avoided looking at grammar.300 and
> your lexer + missing terminator insertion code altogether, to try and
> keep my program completely indepedent.

Okay. In case of conflict, though, grammar.300 is authoritative.

--

Schlingt dreifach einen Kreis um dies! || John Cowan <jco...@reutershealth.com>
Schliesst euer Aug vor heiliger Schau, || http://www.reutershealth.com
Denn er genoss vom Honig-Tau, || http://www.ccil.org/~cowan
Und trank die Milch vom Paradies. -- Coleridge (tr. Politzer)

John Cowan

unread,
Jun 1, 2000, 3:01:37 PM6/1/00
to Bob LeChevalier (lojbab), Lojban List
At 11:39 PM 05/19/2000 +0100, Richard Curnow wrote:

> pu zi gi le masti gi le vensa mi klama le zdani
>
>Unfortunately not! The pu is 'absorbed' into the tense associated with
>the 'gi' construct, so the whole pu..vensa phrase becomes the x1 of klama,
>instead of 'mi', which becomes x2.

I agree that this is a serious problem. A quick-and-dirty fix, involving
no new cmavo and only one new grammar rule, is to add

mod_head_490 term_set_85

as a new alternative for modifier_84. Then the above example can be written:

pu nu'i zigi le masti gi le vensa nu'u mi klama le zdani

This only extends the language and does not prohibit anything that existed
before. Yacc does not complain.

Reply all
Reply to author
Forward
0 new messages