WikiDiscuss

WikiDiscuss


BPFK Section: Erasures

posts: 14214
Use this thread to discuss the BPFK Section: Erasures page.
posts: 14214

Ready for bitch^H^H^H^H^Hvoting.

-Robin

posts: 1912

Erases the last Lojban word, treating certain special cases as a single word. These cases are due to special cmavo which are identified in their definitions as changing other words into words of the pseudo selma'o any-string or any-word.


I would list the cases instead of saying how they are identified in their definitions, otherwise someone reading this will have to read every definition to find out what they are. If I'm not forgetting anything, there are only five of them anyway: zo-quoted word, lo'u-quoted string of words, zoi-quoted foreign text, bu-lerfu and zei-lujvo.

mi tatpi cisma bu si frumu bu

I'm tired. :-) uh :-(

This is an example only; no-one actually uses bu that way.


But even if no-one uses it that way, the English translation is incorrect.
It should be something like: "I'm tired of :-), I mean, of :-(".
Lerfu are pronouns, not indicators.

"sa" erases the preceding text back until it sees a word of the same selma'o as the word that follows it; this earlier word is also erased. If you have no idea what a selma'o is, read "the same word" for "the same selma'o".


I wouldn't use this chatty style in the definition. If you have no idea what a selma'o is, look it up under "selma'o", but this advice need not be given in the definition of sa. Besides, using the same word, may get you the wrong behaviour.

Words whose selma'o has been changed to any-word or any-string by certain other special cmavo are invisible to "sa" for purposes of deciding what to erase; they are erased as any other word would be.


Again, listing the "certain other special cmavo" would not take long, and it would make the definition more complete. Those certain other special cmavo are zo, ZOI, lo'u, bu and zei.

Maybe something like: words quoted with zo, ZOI or lo'u, or part of a bu- or zei- compound are invisible to "sa" for purposes of...

Multiple "sa" before a word erase back to successively further instances of the same selma'o, one for each "sa".


I don't agree with this rule. It is usually hard enough to locate the last member of the selma'o of the following word. Counting sa's and then counting back for members of a given selma'o seems totally undoable. sa sa should behave as expected: looks for previous occurence of sa, doesn't find any and therefore deletes to the beginning of the text, like sa si and sa su.

To completely destroy an utterance, use "sa", the word that started the utterance, and "si", although this won't work if the word in question will quote "si" (such as "zo").


It also won't work if the word in question has been used more than once, or if another word of the same selma'o has been used. You've already said that sasi and sasu will completely destroy an utterance, so there is no need for this sentence.

Examples of sa Usage All translations of "si" into English are approximate at best.


s/si/sa

su (SU) Erase previous discourse. Erases all words back to the beginning of the discourse or text. More precisely, "su" erases back to the previous word of selma'o NIhO, LU, TUhE, or TO.


I disagree with this rule. I think {su} should always erase to the beginning of the utterance. We already have {sa ni'o}, {sa lu}, {sa tu'e}, {sa to} for the other function, the extra syllable or two is hardly a problem for this rather drastic and hopefully rare function. Having to keep these four selmaho especially identified for su-purposes is an unnecessary burden on the listener.

mi'e xorxes

posts: 14214

On Sun, Jan 02, 2005 at 10:28:35AM -0800, wikidiscuss@lojban.org
wrote:
> Re: BPFK Section: Erasures
>

>
Erases the last Lojban word, treating certain special cases as a

> single word. These cases are due to special cmavo which are
> identified in their definitions as changing other words into words

> of the pseudo selma'o any-string or any-word.

>
> I would list the cases instead of saying how they are identified
> in their definitions, otherwise someone reading this will have to
> read every definition to find out what they are. If I'm not
> forgetting anything, there are only five of them anyway: zo-quoted
> word, lo'u-quoted string of words, zoi-quoted foreign text,
> bu-lerfu and zei-lujvo.

Done.

>
mi tatpi cisma bu si frumu bu

> I'm tired. :-) uh :-(

> This is an example only; no-one actually uses bu that way.

>
> But even if no-one uses it that way, the English translation is
> incorrect. It should be something like: "I'm tired of :-), I mean,
> of :-(". Lerfu are pronouns, not indicators.

Shouldn't you need "zo" or "me'o" for your translation? Seems like
what I have is actually "I'm tired of the thing represented by
:-)", whereas the translation I have si "mi tatpi sei cisma bu" or
something.

This actually lead to me finding a bug. camxes does not accept:

mi tatpi me'o cisma bu si re

when I'm pretty sure it should do so.

In fact, si erases bu in general without erasing the word before it.
I'm pretty sure that didn't used to be the case. I wonder if the
morphology changes broke something. No, guess not: SI erases "zo da
bu" but only the "bu" in "da bu".

<sigh>

>
"sa" erases the preceding text back until it sees a word of the

> same selma'o as the word that follows it; this earlier word is
> also erased. If you have no idea what a selma'o is, read "the same

> word" for "the same selma'o".

>
> I wouldn't use this chatty style in the definition. If you have no
> idea what a selma'o is, look it up under "selma'o", but this
> advice need not be given in the definition of sa.

I've had several complaints about SA's use of selma'o being too
complicated because newbies don't know what that is.

> Besides, using the same word, may get you the wrong behaviour.

When, exactly?

>
Words whose selma'o has been changed to any-word or any-string by

> certain other special cmavo are invisible to "sa" for purposes of
> deciding what to erase; they are erased as any other word would

> be.

>
> Again, listing the "certain other special cmavo" would not take
> long, and it would make the definition more complete. Those
> certain other special cmavo are zo, ZOI, lo'u, bu and zei.

Done. It's only BU and ZEI in this case, though.

>
To completely destroy an utterance, use "sa", the word that

> started the utterance, and "si", although this won't work if the

> word in question will quote "si" (such as "zo").

>
> It also won't work if the word in question has been used more than
> once, or if another word of the same selma'o has been used. You've
> already said that sasi and sasu will completely destroy an
> utterance, so there is no need for this sentence.

That sentence predates that other sentence. Fixed.

>
Examples of sa Usage > All translations of "si" into English are approximate at best.

>
> s/si/sa

Done.

-Robin


posts: 14214

On Sun, Jan 02, 2005 at 10:28:35AM -0800, wikidiscuss@lojban.org wrote:

>
Multiple "sa" before a word erase back to successively further > instances of the same selma'o, one for each "sa".

>
> I don't agree with this rule. It is usually hard enough to locate
> the last member of the selma'o of the following word. Counting
> sa's and then counting back for members of a given selma'o seems
> totally undoable. sa sa should behave as expected: looks for
> previous occurence of sa, doesn't find any and therefore deletes
> to the beginning of the text, like sa si and sa su.

Disagreement noted. I don't much care, although I mildly prefer it
as is.

Others?

-Robin


posts: 14214

On Sun, Jan 02, 2005 at 10:28:35AM -0800, wikidiscuss@lojban.org wrote:

>
su (SU) Erase previous discourse. Erases all words back to the

> beginning of the discourse or text. More precisely, "su" erases

> back to the previous word of selma'o NIhO, LU, TUhE, or TO.

>
> I disagree with this rule.

It's from the CLL.

> I think {su} should always erase to the beginning of the
> utterance.

s/utterance/text/, no?

> We already have {sa ni'o}, {sa lu}, {sa tu'e}, {sa to} for the
> other function, the extra syllable or two is hardly a problem for
> this rather drastic and hopefully rare function. Having to keep
> these four selmaho especially identified for su-purposes is an
> unnecessary burden on the listener.

Umm, except that {su} as you've described it is redundant with {sa
si} and {sa su}.

So there's redundancy either way.

-Robin


Robin Lee Powell scripsit:
> On Sun, Jan 02, 2005 at 10:28:35AM -0800, wikidiscuss@lojban.org wrote:

> >
Multiple "sa" before a word erase back to successively further > > instances of the same selma'o, one for each "sa".

> >
> > I don't agree with this rule. It is usually hard enough to locate
> > the last member of the selma'o of the following word. Counting
> > sa's and then counting back for members of a given selma'o seems
> > totally undoable. sa sa should behave as expected: looks for
> > previous occurence of sa, doesn't find any and therefore deletes
> > to the beginning of the text, like sa si and sa su.

On this view, the second "sa" is effectively quoted rather than interpreted,
but normally the word after "sa" is to be interpreted.

> Disagreement noted. I don't much care, although I mildly prefer it
> as is.

I think the most reasonable thing is either the existing rule or to ignore
additional sa's.

--
A mosquito cried out in his pain, John Cowan
"A chemist has poisoned my brain!" http://www.ccil.org/~cowan
The cause of his sorrow http://www.reutershealth.com
Was para-dichloro- jcowan@reutershealth.com
Diphenyltrichloroethane. (aka DDT)


posts: 14214

> > Seems like what I have is actually "I'm tired of the thing
> > represented by :-)", whereas the translation I have si "mi tatpi
> > sei cisma bu" or something.
>
> {sei me cisma bu} maybe. {sei} takes a selbri.

Oh, yeah. I really, really hate that. {sei} should work like {to}.

> > >
"sa" erases the preceding text back until it sees a word of

> > > the same selma'o as the word that follows it; this earlier
> > > word is also erased. If you have no idea what a selma'o is,

> > > read "the same word" for "the same selma'o".

> > >
> > > I wouldn't use this chatty style in the definition. If you
> > > have no idea what a selma'o is, look it up under "selma'o",
> > > but this advice need not be given in the definition of sa.
> >
> > I've had several complaints about SA's use of selma'o being too
> > complicated because newbies don't know what that is.
>
> So are you proposing to change SA to something else? I wouldn't
> mind.

I'm not, no. I'm just explaining why that note is there. People
wanted to change it to "erases to the previous same word", which I
strongly dislike.

> > > Besides, using the same word, may get you the wrong behaviour.
> >
> > When, exactly?
>
> {da de di sa da} for example.

<nod> Point.

Removed.

-Robin


posts: 14214

On Wed, Jan 05, 2005 at 01:29:01PM -0800, Jorge Llamb?as wrote:
> --- Robin Lee Powell wrote:
> > On Sun, Jan 02, 2005 at 10:28:35AM -0800, wikidiscuss@lojban.org
> > wrote:
> > > I think {su} should always erase to the beginning of the
> > > utterance.
> >
> > s/utterance/text/, no?
>
> Is there a difference?

Not really, no.

> > > We already have {sa ni'o}, {sa lu}, {sa tu'e}, {sa to} for the
> > > other function, the extra syllable or two is hardly a problem
> > > for this rather drastic and hopefully rare function. Having to
> > > keep these four selmaho especially identified for su-purposes
> > > is an unnecessary burden on the listener.
> >
> > Umm, except that {su} as you've described it is redundant with
> > {sa si} and {sa su}.
> >
> > So there's redundancy either way.
>
> Yes, I'm not saying: It is redundant, therefore I suggest we
> change it.
>
> I'm saying: It is too complicated, therefore I suggest we change
> it, and since it is redundant anyway, we don't lose anything.

I think having to use sa confuses people, but I don't much care.

Other opinions?

-Robin


Jorge Llamb��)B�as scripsit:

> I would agree with a rule for ignoring additional sa's if the
> rule for sa in general was to do nothing when it doesn't find
> any previous word of the same selmaho of the following word.
> Then sa sa wouldn't be an exception.

Sounds good to me.

--
John Cowan jcowan@reutershealth.com http://www.ccil.org/~cowan
Does anybody want any flotsam? / I've gotsam.
Does anybody want any jetsam? / I can getsam.
--Ogden Nash, No Doctors Today, Thank You


posts: 14214

On Wed, Jan 05, 2005 at 04:49:52PM -0500, John Cowan wrote:
> Jorge Llamb??????)B???as scripsit:
>
> > I would agree with a rule for ignoring additional sa's if the
> > rule for sa in general was to do nothing when it doesn't find
> > any previous word of the same selmaho of the following word.
> > Then sa sa wouldn't be an exception.
>
> Sounds good to me.

Not to me. Not even a bit.

People should be able to put chat logs and such into a parser
line-by-line and get useful results out. Given:

foo: mi klama
bar: klama ma
foo: lo zdani be la fred .y. sa do klama lo zdani be mi

The result should *not* be:

mi klama

klama ma

lo zdani be la fred .y. do klama lo zdani be mi

Not least because the last line is now *ungrammatical*.

-Robin

--
http://www.digitalkingdom.org/~rlpowell/ *** http://www.lojban.org/
Reason #237 To Learn Lojban: "Homonyms: Their Grate!"
Proud Supporter of the Singularity Institute - http://singinst.org/


posts: 14214

On Wed, Jan 05, 2005 at 02:54:26PM -0800, Jorge Llamb?as wrote:
> This is just to say that what happens with incomplete chat-log
> lines is not really what we should be looking at to decide the
> best behaviour of sa.

As greater than half of my test suite is "incomplete" chat logs,
my opinion differs.

-Robin, who has *no* interest in reviewing > 5 K lines of input

  • again*.



posts: 14214

On Wed, Jan 05, 2005 at 03:12:11PM -0800, Jorge Llamb?as wrote:
> --- Robin Lee Powell wrote:
> > On Wed, Jan 05, 2005 at 02:54:26PM -0800, Jorge Llamb?as wrote:
> > > This is just to say that what happens with incomplete chat-log
> > > lines is not really what we should be looking at to decide the
> > > best behaviour of sa.
> >
> > As greater than half of my test suite is "incomplete" chat logs,
> > my opinion differs.
> >
> > -Robin, who has *no* interest in reviewing > 5 K lines of input
> > *again*.
>
> How many lines of the test suite could possibly contain {sa}.

Good point.

460.

-Robin


posts: 152

> > I wouldn't use this chatty style in the definition. If you have no
> > idea what a selma'o is, look it up under "selma'o", but this
> > advice need not be given in the definition of sa.
>
> I've had several complaints about SA's use of selma'o being too
> complicated because newbies don't know what that is.

I think the complaint is that it's a very unnatural and counterintuitive rule.
It's not just newbies who shouldn't have to know what selma'o are; eventually,
no user of the language should have to know what selma'o are. They're a tool
for talking about the language, not part of the language.

--
Rob Speer



On Wednesday 05 January 2005 21:37, Rob Speer wrote:
> I think the complaint is that it's a very unnatural and counterintuitive
> rule. It's not just newbies who shouldn't have to know what selma'o are;
> eventually, no user of the language should have to know what selma'o are.
> They're a tool for talking about the language, not part of the language.

Lots of English speakers don't know what an animate noun is, but if you say
"The cow was found by a stream by a farmer", they know who found the cow.

phma
--
My monthly periods happen once per year.
-Les Perles de la médecine


posts: 152

On Wed, Jan 05, 2005 at 10:10:52PM -0500, Pierre Abbat wrote:
> On Wednesday 05 January 2005 21:37, Rob Speer wrote:
> > I think the complaint is that it's a very unnatural and counterintuitive
> > rule. It's not just newbies who shouldn't have to know what selma'o are;
> > eventually, no user of the language should have to know what selma'o are.
> > They're a tool for talking about the language, not part of the language.
>
> Lots of English speakers don't know what an animate noun is, but if you say
> "The cow was found by a stream by a farmer", they know who found the cow.

That's because they can use some relatively straightforward semantics to get
the most likely interpretation of the sentence. And English never claims to be
unambiguous, so there's no problem if they don't know the "rule" about animate
nouns precisely enough.

How do you define "the same selma'o" without requiring memorization of a whole
bunch of lists of cmavo? It's not "can be interchanged grammatically in any
context", because we know there are distinct selma'o that are grammatically
equivalent.
--
Rob Speer



posts: 14214

On Wed, Jan 05, 2005 at 11:01:22PM -0500, Rob Speer wrote:
> How do you define "the same selma'o" without requiring
> memorization of a whole bunch of lists of cmavo? It's not "can be
> interchanged grammatically in any context", because we know there
> are distinct selma'o that are grammatically equivalent.

Not if xorxes has his way. :-)

It's worth noting that switching to a same-world rule would
dramatically increase the parser's context sensitivity, and possibly
slow it down quite a bit. This isn't an argument as such, just
mentioning.

-Robin


On Thursday 06 January 2005 09:02, Jorge "Llambías" wrote:
> I would certainly like to reduce the number of selma'o,
> and not only those that are identical but for the SA rule
> (TAhE and ZAhO, UI and CAI, DAhO and FUhO, and I think
> that's it). I would put UI, CAI, NAI, FUhE, DAhO, FUhO,
> RAhO and GAhO all in the same selma'o, for example,
> and the same for BAI, CAhA, CUhE, KI, ZI, PU, VA, ZEhA,
> VEhA and VIhA (see
> <http://www.lojban.org/tiki/tiki-index.php?page=Internal+grammar+of+tags>).

I think RAhO and GAhO should be kept separate. RAhO occurs only after GOhA,
and GAhO occurs only next to BIhI. What would make sense is allowing RAhO
after brivla and considering all GOhA as brivla, and allowing BIhI with only
one GAhO.

phma

--
li ze te'a ci vu'u ci bi'e te'a mu du
li ci su'i ze te'a mu bi'e vu'u ci


posts: 1912


Instead of looking at the selmaho of the following word
and then erasing back until another member of that selmaho
is found, SA could look at the following major construct and
erase back until another start of such construct is found.

The main two constructs involved would be "term" and "bridi-tail",
since those would have to be the most common uses of SA:

mi klama le zarci sa le zdani
I go to the market, I mean, to the house.

(That one doesn't change.)

mi viska la djan sa le bruna be la djan
I saw John, I mean, John's brother.

mi viska le bruna be la djan sa tirna by
I saw John's brother, I mean, heard him.

(Whether pronouns will work like this is a separate issue.
The point here is that tirna replaces {viska}, not {bruna},
because it is a bridi-tail.)

The construct to be erased need not have been completed,
this will also work:

mi viska le sa la djan
I saw the, I mean, John.

The constructs that SA would detect are:

term
bridi-tail
sentence
relative-clause
links (be)
linkargs (bei)
joik-ek (sumti connective)
gihek (bridi-tail connective)

(and I suppose operands and operators within MEX)

I would have SA simply ignore free modifiers and indicators.

This would require some 10 rules for SA instead of the
current 120, and people would not need to learn what
word belongs in what selmaho. The substitution of a term with
a term, a selbri with a selbri, a relative clause with a relative
clause and so on is also intuitive enough that people won't
really have to think in terms of the grammatical construct
in order to apply the rule.

Opinions?

mu'o mi'e xorxes

posts: 14214

On Mon, Jan 10, 2005 at 10:42:51AM -0800, wikidiscuss@lojban.org
wrote:
> Instead of looking at the selmaho of the following word and then
> erasing back until another member of that selmaho is found, SA
> could look at the following major construct and erase back until
> another start of such construct is found.

Any opinions at all?

-Robin


posts: 149

Robin Lee Powell scripsit:
> On Mon, Jan 10, 2005 at 10:42:51AM -0800, wikidiscuss@lojban.org
> wrote:
> > Instead of looking at the selmaho of the following word and then
> > erasing back until another member of that selmaho is found, SA
> > could look at the following major construct and erase back until
> > another start of such construct is found.
>
> Any opinions at all?

The proof of the pudding's in the eating, but I don't think this is
ridiculous. Let's see a working PEG grammar for it.

--
John Cowan www.reutershealth.com www.ccil.org/~cowan cowan@ccil.org
'Tis the Linux rebellion / Let coders take their place,
The Linux-nationale / Shall Microsoft outpace,
We can write better programs / Our CPUs won't stall,
So raise the penguin banner of / The Linux-nationale. --Greg Baker


posts: 381

In a message dated 2005-01-11 2:12:53 AM Eastern Standard Time,
rlpowell@digitalkingdom.org writes:


>
> On Mon, Jan 10, 2005 at 10:42:51AM -0800, wikidiscuss@lojban.org
> wrote:
> > Instead of looking at the selmaho of the following word and then
> > erasing back until another member of that selmaho is found, SA
> > could look at the following major construct and erase back until
> > another start of such construct is found.
>
> Any opinions at all?
>
> -Robin
>
>
>
>
>

I think it sounds GREAT!
stevo

posts: 152

On Mon, Jan 10, 2005 at 11:12:25PM -0800, Robin Lee Powell wrote:
> On Mon, Jan 10, 2005 at 10:42:51AM -0800, wikidiscuss@lojban.org
> wrote:
> > Instead of looking at the selmaho of the following word and then
> > erasing back until another member of that selmaho is found, SA
> > could look at the following major construct and erase back until
> > another start of such construct is found.
>
> Any opinions at all?

I like it, if it works. It seems like it should be a bit fragile, though - what
happens if there's a SA in the next phrase too?
--
Rob Speer



posts: 14214

On Mon, Jan 10, 2005 at 10:42:51AM -0800, wikidiscuss@lojban.org wrote:
> mi viska le bruna be la djan sa tirna by
>
> I saw John's brother, I mean, heard him.
>
> (Whether pronouns will work like this is a separate issue.
> The point here is that tirna replaces {viska}, not {bruna},
> because it is a bridi-tail.)

Erm, isn't bruna too?

Oh, no, it's a sumti-tail. OK. I find that a little bizarre (that
a BRIVLA wouldn't SA back to the previous BRIVLA) but of course I'm
used to the old way. :-)

> I would have SA simply ignore free modifiers and indicators.

By which you mean that it ignores them looking forward for things to
match?

mi viska le bruna be la djan sa sei mi nelci se'u le broda

==

mi viska le bruna be le broda

?

Seems that if it ignores them in searching forward for things to
match is equivalent to erasing them.

-Robin

--
http://www.digitalkingdom.org/~rlpowell/ *** http://www.lojban.org/
Reason #237 To Learn Lojban: "Homonyms: Their Grate!"
Proud Supporter of the Singularity Institute - http://singinst.org/


posts: 14214

On Tue, Jan 11, 2005 at 01:57:39PM -0500, Rob Speer wrote:
> On Mon, Jan 10, 2005 at 11:12:25PM -0800, Robin Lee Powell wrote:
> > On Mon, Jan 10, 2005 at 10:42:51AM -0800, wikidiscuss@lojban.org
> > wrote:
> > > Instead of looking at the selmaho of the following word and
> > > then erasing back until another member of that selmaho is
> > > found, SA could look at the following major construct and
> > > erase back until another start of such construct is found.
> >
> > Any opinions at all?
>
> I like it, if it works. It seems like it should be a bit fragile,
> though - what happens if there's a SA in the next phrase too?

They collapse sequentially, in LTR order. Can you give an example
of what you're worried about?

-Robin

--
http://www.digitalkingdom.org/~rlpowell/ *** http://www.lojban.org/
Reason #237 To Learn Lojban: "Homonyms: Their Grate!"
Proud Supporter of the Singularity Institute - http://singinst.org/


posts: 152

On Tue, Jan 11, 2005 at 11:06:46AM -0800, Robin Lee Powell wrote:
> They collapse sequentially, in LTR order. Can you give an example
> of what you're worried about?

Do you mean right-to-left? I suppose that would work. (Left-to-right wouldn't
work, because you don't know what the first SA does until you know how the
phrase containing the second SA parses.)
--
Rob Speer



posts: 14214

On Tue, Jan 11, 2005 at 02:15:41PM -0500, Rob Speer wrote:
> On Tue, Jan 11, 2005 at 11:06:46AM -0800, Robin Lee Powell wrote:
> > They collapse sequentially, in LTR order. Can you give an
> > example of what you're worried about?
>
> Do you mean right-to-left? I suppose that would work.

No, I did not.

> (Left-to-right wouldn't work, because you don't know what the
> first SA does until you know how the phrase containing the second
> SA parses.)

Example, please.

-Robin


posts: 1912

> I think RAhO and GAhO should be kept separate. RAhO occurs only after GOhA,
> and GAhO occurs only next to BIhI.

That's how they are defined now, yes, but I've had occasion to use both
in other positions. {ga'o} could have a more general sense of "included"
and {ke'i} "excluded", so for example I could say:

mi viska ro le prenu e la djan ga'o
I saw everybody, including John.

{ra'o} need not be attached to the selbri in order to update pronouns.
You may want to say {ta e ta ra'o e ta ra'o} for "that one, and that one
(another one) and that one (another one). {... ta ... ije ... ta ra'o
nai ...}, "... that one ... that one (the same one) ...", and I'm sure
there are other uses.

> What would make sense is allowing RAhO
> after brivla and considering all GOhA as brivla, and allowing BIhI with only
> one GAhO.

That would be a first step in the right direction.

mu'o mi'e xorxes




__
Do you Yahoo!?
Meet the all-new My Yahoo! - Try it today!
http://my.yahoo.com




posts: 1912

> On Wed, Jan 05, 2005 at 11:01:22PM -0500, Rob Speer wrote:
> > How do you define "the same selma'o" without requiring
> > memorization of a whole bunch of lists of cmavo? It's not "can be
> > interchanged grammatically in any context", because we know there
> > are distinct selma'o that are grammatically equivalent.
>
> Not if xorxes has his way. :-)

No two selma'o are interchangeable with the current grammar,
precisely because the SA rule makes sure they are not
interchangeable.

I would certainly like to reduce the number of selma'o,
and not only those that are identical but for the SA rule
(TAhE and ZAhO, UI and CAI, DAhO and FUhO, and I think
that's it). I would put UI, CAI, NAI, FUhE, DAhO, FUhO,
RAhO and GAhO all in the same selma'o, for example,
and the same for BAI, CAhA, CUhE, KI, ZI, PU, VA, ZEhA,
VEhA and VIhA (see
<http://www.lojban.org/tiki/tiki-index.php?page=Internal+grammar+of+tags>).

> It's worth noting that switching to a same-world rule would
> dramatically increase the parser's context sensitivity, and possibly
> slow it down quite a bit. This isn't an argument as such, just
> mentioning.

I don't think same-word would be a good rule either. I would
rather have something like: SA+selbri erases from last beginning
of bridi-tail, SA+term erases from last beginning of term,
SA+relative-clause erases from last beginning of relative-clause,
SA+(I/NIhO) erases from last beginning of new sentence, and that's
about it, just the major structures.

mu'o mi'e xorxes






__
Do you Yahoo!?
The all-new My Yahoo! - Get yours free!
http://my.yahoo.com




posts: 1912

> On this view, the second "sa" is effectively quoted rather than interpreted,
> but normally the word after "sa" is to be interpreted.

I'm saying it should be interpreted like any other word:

When you get to the first sa, you look at the next word
and erase everything till the previous occurrence of a
word of that selmaho or to the beginning of text if no
such word is found. (Since you won't find another sa,
that means you will erase to the beginning of text.)

Now you do the same for the next sa. Since it is now
at the beginning of text, it has nothing else to erase.

It is the same interpretation for every word.
The current implementation interprets the second sa
differently than other words.

> I think the most reasonable thing is either the existing rule or to ignore
> additional sa's.

I would agree with a rule for ignoring additional sa's if the
rule for sa in general was to do nothing when it doesn't find
any previous word of the same selmaho of the following word.
Then sa sa wouldn't be an exception.

mu'o mi'e xorxes





__
Do you Yahoo!?
The all-new My Yahoo! - Get yours free!
http://my.yahoo.com




posts: 1912

> People should be able to put chat logs and such into a parser
> line-by-line and get useful results out. Given:
>
> foo: mi klama
> bar: klama ma
> foo: lo zdani be la fred .y. sa do klama lo zdani be mi
>
> The result should *not* be:
>
> mi klama
>
> klama ma
>
> lo zdani be la fred .y. do klama lo zdani be mi
>
> Not least because the last line is now *ungrammatical*.

What about:

foo: la djan cu klama ma
bar: le zdani
foo: oi se'i sa cu dunda ma

The last line by itself will be ungrammatical no matter what
{sa} does.

This is just to say that what happens with incomplete chat-log
lines is not really what we should be looking at to decide the
best behaviour of sa.

mu'o mi'e xorxes





__
Do you Yahoo!?
The all-new My Yahoo! - What will yours do?
http://my.yahoo.com


posts: 1912

> On Wed, Jan 05, 2005 at 02:54:26PM -0800, Jorge Llamb?as wrote:
> > This is just to say that what happens with incomplete chat-log
> > lines is not really what we should be looking at to decide the
> > best behaviour of sa.
>
> As greater than half of my test suite is "incomplete" chat logs,
> my opinion differs.
>
> -Robin, who has *no* interest in reviewing > 5 K lines of input
> *again*.

How many lines of the test suite could possibly contain {sa}.

mu'o mi'e xorxes




__
Do you Yahoo!?
Yahoo! Mail - Easier than ever with enhanced search. Learn more.
http://info.mail.yahoo.com/mail_250


Re: BPFK Section: Erasures

Erases the last Lojban word, treating certain special cases as a single word. These cases are due to special cmavo which are identified in their definitions as changing other words into words of the pseudo selma'o any-string or any-word.


I would list the cases instead of saying how they are identified in their definitions, otherwise someone reading this will have to read every definition to find out what they are. If I'm not forgetting anything, there are only five of them anyway: zo-quoted word, lo'u-quoted string of words, zoi-quoted foreign text, bu-lerfu and zei-lujvo.

mi tatpi cisma bu si frumu bu

I'm tired. :-) uh :-(

This is an example only; no-one actually uses bu that way.


But even if no-one uses it that way, the English translation is incorrect.
It should be something like: "I'm tired of :-), I mean, of :-(".
Lerfu are pronouns, not indicators.

"sa" erases the preceding text back until it sees a word of the same selma'o as the word that follows it; this earlier word is also erased. If you have no idea what a selma'o is, read "the same word" for "the same selma'o".


I wouldn't use this chatty style in the definition. If you have no idea what a selma'o is, look it up under "selma'o", but this advice need not be given in the definition of sa. Besides, using the same word, may get you the wrong behaviour.

Words whose selma'o has been changed to any-word or any-string by certain other special cmavo are invisible to "sa" for purposes of deciding what to erase; they are erased as any other word would be.


Again, listing the "certain other special cmavo" would not take long, and it would make the definition more complete. Those certain other special cmavo are zo, ZOI, lo'u, bu and zei.

Maybe something like: words quoted with zo, ZOI or lo'u, or part of a bu- or zei- compound are invisible to "sa" for purposes of...

Multiple "sa" before a word erase back to successively further instances of the same selma'o, one for each "sa".


I don't agree with this rule. It is usually hard enough to locate the last member of the selma'o of the following word. Counting sa's and then counting back for members of a given selma'o seems totally undoable. sa sa should behave as expected: looks for previous occurence of sa, doesn't find any and therefore deletes to the beginning of the text, like sa si and sa su.

To completely destroy an utterance, use "sa", the word that started the utterance, and "si", although this won't work if the word in question will quote "si" (such as "zo").


It also won't work if the word in question has been used more than once, or if another word of the same selma'o has been used. You've already said that sasi and sasu will completely destroy an utterance, so there is no need for this sentence.

Examples of sa Usage All translations of "si" into English are approximate at best.


s/si/sa

su (SU) Erase previous discourse. Erases all words back to the beginning of the discourse or text. More precisely, "su" erases back to the previous word of selma'o NIhO, LU, TUhE, or TO.


I disagree with this rule. I think {su} should always erase to the beginning of the utterance. We already have {sa ni'o}, {sa lu}, {sa tu'e}, {sa to} for the other function, the extra syllable or two is hardly a problem for this rather drastic and hopefully rare function. Having to keep these four selmaho especially identified for su-purposes is an unnecessary burden on the listener.

mi'e xorxes





posts: 14214

On Sun, Jan 02, 2005 at 10:28:35AM -0800, wikidiscuss@lojban.org
wrote:
> Re: BPFK Section: Erasures
>

>
Erases the last Lojban word, treating certain special cases as a

> single word. These cases are due to special cmavo which are
> identified in their definitions as changing other words into words

> of the pseudo selma'o any-string or any-word.

>
> I would list the cases instead of saying how they are identified
> in their definitions, otherwise someone reading this will have to
> read every definition to find out what they are. If I'm not
> forgetting anything, there are only five of them anyway: zo-quoted
> word, lo'u-quoted string of words, zoi-quoted foreign text,
> bu-lerfu and zei-lujvo.

Done.

>
mi tatpi cisma bu si frumu bu

> I'm tired. :-) uh :-(

> This is an example only; no-one actually uses bu that way.

>
> But even if no-one uses it that way, the English translation is
> incorrect. It should be something like: "I'm tired of :-), I mean,
> of :-(". Lerfu are pronouns, not indicators.

Shouldn't you need "zo" or "me'o" for your translation? Seems like
what I have is actually "I'm tired of the thing represented by
:-)", whereas the translation I have si "mi tatpi sei cisma bu" or
something.

This actually lead to me finding a bug. camxes does not accept:

mi tatpi me'o cisma bu si re

when I'm pretty sure it should do so.

In fact, si erases bu in general without erasing the word before it.
I'm pretty sure that didn't used to be the case. I wonder if the
morphology changes broke something. No, guess not: SI erases "zo da
bu" but only the "bu" in "da bu".

<sigh>

>
"sa" erases the preceding text back until it sees a word of the

> same selma'o as the word that follows it; this earlier word is
> also erased. If you have no idea what a selma'o is, read "the same

> word" for "the same selma'o".

>
> I wouldn't use this chatty style in the definition. If you have no
> idea what a selma'o is, look it up under "selma'o", but this
> advice need not be given in the definition of sa.

I've had several complaints about SA's use of selma'o being too
complicated because newbies don't know what that is.

> Besides, using the same word, may get you the wrong behaviour.

When, exactly?

>
Words whose selma'o has been changed to any-word or any-string by

> certain other special cmavo are invisible to "sa" for purposes of
> deciding what to erase; they are erased as any other word would

> be.

>
> Again, listing the "certain other special cmavo" would not take
> long, and it would make the definition more complete. Those
> certain other special cmavo are zo, ZOI, lo'u, bu and zei.

Done. It's only BU and ZEI in this case, though.

>
To completely destroy an utterance, use "sa", the word that

> started the utterance, and "si", although this won't work if the

> word in question will quote "si" (such as "zo").

>
> It also won't work if the word in question has been used more than
> once, or if another word of the same selma'o has been used. You've
> already said that sasi and sasu will completely destroy an
> utterance, so there is no need for this sentence.

That sentence predates that other sentence. Fixed.

>
Examples of sa Usage > All translations of "si" into English are approximate at best.

>
> s/si/sa

Done.

-Robin

posts: 1912

> >
mi tatpi cisma bu si frumu bu

> > I'm tired. :-) uh :-(

> > This is an example only; no-one actually uses bu that way.

> >
> > But even if no-one uses it that way, the English translation is
> > incorrect. It should be something like: "I'm tired of :-), I mean,
> > of :-(". Lerfu are pronouns, not indicators.
>
> Shouldn't you need "zo" or "me'o" for your translation?

It depends how you interpret "I'm tired of :-)".
Does ":-)" there refer to itself, or is it used as
a symbol? If I had quoted the happy-face, then I
agree it would need zo. In any case, it is not standard
English, but using {cisma bu} as a pronoun is not
common in Lojban either.

> Seems like
> what I have is actually "I'm tired of the thing represented by
> :-)", whereas the translation I have si "mi tatpi sei cisma bu" or
> something.

{sei me cisma bu} maybe. {sei} takes a selbri.

> This actually lead to me finding a bug. camxes does not accept:
>
> mi tatpi me'o cisma bu si re
>
> when I'm pretty sure it should do so.

I'll save that for later.

> >
"sa" erases the preceding text back until it sees a word of the

> > same selma'o as the word that follows it; this earlier word is
> > also erased. If you have no idea what a selma'o is, read "the same

> > word" for "the same selma'o".

> >
> > I wouldn't use this chatty style in the definition. If you have no
> > idea what a selma'o is, look it up under "selma'o", but this
> > advice need not be given in the definition of sa.
>
> I've had several complaints about SA's use of selma'o being too
> complicated because newbies don't know what that is.

So are you proposing to change SA to something else?
I wouldn't mind.

> > Besides, using the same word, may get you the wrong behaviour.
>
> When, exactly?

{da de di sa da} for example.

mu'o mi'e xorxes




__
Do you Yahoo!?
Take Yahoo! Mail with you! Get it on your mobile phone.
http://mobile.yahoo.com/maildemo


posts: 1912

> On Sun, Jan 02, 2005 at 10:28:35AM -0800, wikidiscuss@lojban.org wrote:
> > I think {su} should always erase to the beginning of the
> > utterance.
>
> s/utterance/text/, no?

Is there a difference?

text can be confused with the text rule. Erasing the current
"text" in the rule sense is more or less what SU does now,
because LU, TUhE and TO take a text (or something like it).
(NIhO not quite.)

> > We already have {sa ni'o}, {sa lu}, {sa tu'e}, {sa to} for the
> > other function, the extra syllable or two is hardly a problem for
> > this rather drastic and hopefully rare function. Having to keep
> > these four selmaho especially identified for su-purposes is an
> > unnecessary burden on the listener.
>
> Umm, except that {su} as you've described it is redundant with {sa
> si} and {sa su}.
>
> So there's redundancy either way.

Yes, I'm not saying: It is redundant, therefore I suggest we
change it.

I'm saying: It is too complicated, therefore I suggest we change
it, and since it is redundant anyway, we don't lose anything.

mu'o mi'e xorxes




__
Do you Yahoo!?
The all-new My Yahoo! - Get yours free!
http://my.yahoo.com




Re: BPFK Section: Erasures

Instead of looking at the selmaho of the following word
and then erasing back until another member of that selmaho
is found, SA could look at the following major construct and
erase back until another start of such construct is found.

The main two constructs involved would be "term" and "bridi-tail",
since those would have to be the most common uses of SA:

mi klama le zarci sa le zdani
I go to the market, I mean, to the house.

(That one doesn't change.)

mi viska la djan sa le bruna be la djan
I saw John, I mean, John's brother.

mi viska le bruna be la djan sa tirna by
I saw John's brother, I mean, heard him.

(Whether pronouns will work like this is a separate issue.
The point here is that tirna replaces {viska}, not {bruna},
because it is a bridi-tail.)

The construct to be erased need not have been completed,
this will also work:

mi viska le sa la djan
I saw the, I mean, John.

The constructs that SA would detect are:

term
bridi-tail
sentence
relative-clause
links (be)
linkargs (bei)
joik-ek (sumti connective)
gihek (bridi-tail connective)

(and I suppose operands and operators within MEX)

I would have SA simply ignore free modifiers and indicators.

This would require some 10 rules for SA instead of the
current 120, and people would not need to learn what
word belongs in what selmaho. The substitution of a term with
a term, a selbri with a selbri, a relative clause with a relative
clause and so on is also intuitive enough that people won't
really have to think in terms of the grammatical construct
in order to apply the rule.

Opinions?

mu'o mi'e xorxes




posts: 1912

> On Mon, Jan 10, 2005 at 10:42:51AM -0800, wikidiscuss@lojban.org wrote:
> > I would have SA simply ignore free modifiers and indicators.
>
> By which you mean that it ignores them looking forward for things to
> match?

That's what I had in mind, yes.

> mi viska le bruna be la djan sa sei mi nelci se'u le broda
>
> ==
>
> mi viska le bruna be le broda
>
> ?
>
> Seems that if it ignores them in searching forward for things to
> match is equivalent to erasing them.

I was thinking: {mi viska le bruna be sei mi nelci se'u le broda}

mu'o mi'e xorxes




__
Do you Yahoo!?
The all-new My Yahoo! - What will yours do?
http://my.yahoo.com


posts: 14214

On Wed, Jan 12, 2005 at 06:07:36AM -0800, Jorge Llamb?as wrote:
> --- Robin Lee Powell wrote:
> > On Mon, Jan 10, 2005 at 10:42:51AM -0800, wikidiscuss@lojban.org
> > wrote:
> > > I would have SA simply ignore free modifiers and indicators.
> >
> > By which you mean that it ignores them looking forward for
> > things to match?
>
> That's what I had in mind, yes.
>
> > mi viska le bruna be la djan sa sei mi nelci se'u le broda
> >
> > ==
> >
> > mi viska le bruna be le broda
> >
> > ?
> >
> > Seems that if it ignores them in searching forward for things to
> > match is equivalent to erasing them.
>
> I was thinking: {mi viska le bruna be sei mi nelci se'u le broda}

I don't like that so much.

For one thing, it's confusing for the listener, for another, it'll
be hell on the parser (infinite lookahead with complex matching for
every SA it encounters).

Actually, it only needs to look ahead through free* and indicators.
That's not *too* bad.

What do other people think?

Would the above piss you off as a listener?

I want to point out that the main *benefit* to this approach is
things like "mi nelca la broda sa ko'a", i.e. replacement with
wholly different cmavo types that match the same structure.

-Robin



posts: 2388


<rlpowell@digitalkingdom.org> wrote:

> On Wed, Jan 12, 2005 at 06:07:36AM -0800, Jorge
> Llamb?as wrote:
> > --- Robin Lee Powell wrote:
> > > On Mon, Jan 10, 2005 at 10:42:51AM -0800,
> wikidiscuss@lojban.org
> > > wrote:
> > > > I would have SA simply ignore free
> modifiers and indicators.
> > >
> > > By which you mean that it ignores them
> looking forward for
> > > things to match?
> >
> > That's what I had in mind, yes.
> >
> > > mi viska le bruna be la djan sa sei mi
> nelci se'u le broda
> > >
> > > ==
> > >
> > > mi viska le bruna be le broda
> > >
> > > ?
> > >
> > > Seems that if it ignores them in searching
> forward for things to
> > > match is equivalent to erasing them.
> >
> > I was thinking: {mi viska le bruna be sei mi
> nelci se'u le broda}

Since this does not contain an erasure, I don't
see what the point is.

> I don't like that so much.
>
> For one thing, it's confusing for the listener,
> for another, it'll
> be hell on the parser (infinite lookahead with
> complex matching for
> every SA it encounters).
>
> Actually, it only needs to look ahead through
> free* and indicators.
> That's not *too* bad.
>
> What do other people think?
>
> Would the above piss you off as a listener?
>
> I want to point out that the main *benefit* to
> this approach is
> things like "mi nelca la broda sa ko'a", i.e.
> replacement with
> wholly different cmavo types that match the
> same structure.
>
> -Robin
>
>
>
>



On Wed, 12 Jan 2005, Robin Lee Powell wrote:

> On Wed, Jan 12, 2005 at 06:07:36AM -0800, Jorge Llamb?as wrote:
>> --- Robin Lee Powell wrote:
>>> On Mon, Jan 10, 2005 at 10:42:51AM -0800, wikidiscuss@lojban.org
>>> wrote:
>>>> I would have SA simply ignore free modifiers and indicators.
>>>
>>> By which you mean that it ignores them looking forward for
>>> things to match?
>>
>> That's what I had in mind, yes.
>>
>>> mi viska le bruna be la djan sa sei mi nelci se'u le broda
>>>
>>> ==
>>>
>>> mi viska le bruna be le broda
>>>
>>> ?
>>>
>>> Seems that if it ignores them in searching forward for things to
>>> match is equivalent to erasing them.
>>
>> I was thinking: {mi viska le bruna be sei mi nelci se'u le broda}
>
> I don't like that so much.
>
> For one thing, it's confusing for the listener, for another, it'll
> be hell on the parser (infinite lookahead with complex matching for
> every SA it encounters).
>
> Actually, it only needs to look ahead through free* and indicators.
> That's not *too* bad.
>
> What do other people think?
>
> Would the above piss you off as a listener?

It definitely would. Whether or not it should be legal, though, I'm less
certain of. (There are a huge number of ways of pissing off one's audience,
and arbitrarily restricting them can lead to problems.) Personally, I'd
naively read that with the modifiers affecting the sa.

mi viska le bruna be la .djan. sa .u'u la .djein.

Which is more or less equivalent to the modifiers dropping out (along with the
replaced term and the sa). I think whatever gets decided about what they
actually mean, they won't get used much (but should they be banned on those
grounds, and then we lose them when we actually have a good use for them?).

> I want to point out that the main *benefit* to this approach is
> things like "mi nelca la broda sa ko'a", i.e. replacement with
> wholly different cmavo types that match the same structure.

I definitely like the more abstract bridi-tails replacing bridi-tails, sumti
replacing sumti, etc.

One question I'd have is whether there's a way to go further back.

mi viska le burne be la .djan. sa li'o

at this point, I realize that I wanted to say {bruna} instead of {bunre}, but
because I've already started another sumti, it seems I'm stuck with a bunch of
{si}s. Or maybe {sa viska le bruna be la .djan.} would be clearest anyway.

Another question: How far back into tanru does it go when replacing a bridi
tail? That is

mi mutce nelci ko'a sa xebni ko'e

Does the {mutce} remain, or not? I'd be inclined to say not (replace the
entire selbri, you can always add back the parts you need), but I haven't seen
it spelled out yet.
--
Adam Lopresto
http://cec.wustl.edu/~adam/

o \ o / _ o | \ / | o _ \ o / o
/|\ | / \ \o \o | o/ o/ / \ | /|\
/ \ / \ | \ /) | ( \ /o\ / ) | (\ / | / \ / \


posts: 1912


> Personally, I'd
> naively read that with the modifiers affecting the sa.
>
> mi viska le bruna be la .djan. sa .u'u la .djein.
>
> Which is more or less equivalent to the modifiers dropping out (along with
> the
> replaced term and the sa).

I think that's the best interpretation.

>
> One question I'd have is whether there's a way to go further back.
>
> mi viska le burne be la .djan. sa li'o
>
> at this point, I realize that I wanted to say {bruna} instead of {bunre}, but
> because I've already started another sumti, it seems I'm stuck with a bunch
> of
> {si}s. Or maybe {sa viska le bruna be la .djan.} would be clearest anyway.

Yes. Although multiple SA's can probably be made to work,
I don't like it because it involves counting grammar constructs,
which is not something that is practical when speaking
(and when writing you don't need to use SA at all).

> Another question: How far back into tanru does it go when replacing a bridi
> tail? That is
>
> mi mutce nelci ko'a sa xebni ko'e
>
> Does the {mutce} remain, or not? I'd be inclined to say not (replace the
> entire selbri, you can always add back the parts you need), but I haven't
> seen
> it spelled out yet.

That's the idea. The entire bridi-tail gets replaced by the new one.

mu'o mi'e xorxes




__
Do you Yahoo!?
Yahoo! Mail - now with 250MB free storage. Learn more.
http://info.mail.yahoo.com/mail_250


posts: 14214

On Thu, Jan 13, 2005 at 10:00:20AM -0600, Adam D. Lopresto wrote:
> It definitely would. Whether or not it should be legal, though,
> I'm less certain of. (There are a huge number of ways of pissing
> off one's audience, and arbitrarily restricting them can lead to
> problems.) Personally, I'd naively read that with the modifiers
> affecting the sa.

If we're going to allow that sort of thing, that's how I'd read it
as well.

> mi viska le bruna be la .djan. sa .u'u la .djein.

Becomes:

mi viska le bruna be la .djein.

in effect then, yes? Because the .u'u goes away with the sa it
modified? That was my suggestion to xorxes, which he disagreed
with.

> Which is more or less equivalent to the modifiers dropping out
> (along with the replaced term and the sa).

<nod>

> I think whatever gets decided about what they actually mean, they
> won't get used much (but should they be banned on those grounds,
> and then we lose them when we actually have a good use for them?).

Indeed.

> One question I'd have is whether there's a way to go further back.
>
> mi viska le burne be la .djan. sa li'o
>
> at this point, I realize that I wanted to say {bruna} instead of
> {bunre}, but because I've already started another sumti, it seems
> I'm stuck with a bunch of {si}s.

sa da si sa le bruna — the parser doesn't currently like this, but
it should.

The other way is to allow multiple sa, which pretty much everyone
has said they don't like. I kinda think their neat, but not enough
to push on it.

> Another question: How far back into tanru does it go when
> replacing a bridi tail? That is
>
> mi mutce nelci ko'a sa xebni ko'e
>
> Does the {mutce} remain, or not?

Not.

-Robin


posts: 14214

On Thu, Jan 13, 2005 at 08:53:21AM -0800, Jorge Llamb?as wrote:
>
> --- "Adam D. Lopresto" wrote:
> > Personally, I'd naively read that with the modifiers affecting
> > the sa.
> >
> > mi viska le bruna be la .djan. sa .u'u la .djein.
> >
> > Which is more or less equivalent to the modifiers dropping out
> > (along with the replaced term and the sa).
>
> I think that's the best interpretation.

That's what *I* said that you disagreed with, wasn't it?

-Robin


posts: 1912


> > > Which is more or less equivalent to the modifiers dropping out
> > > (along with the replaced term and the sa).
> >
> > I think that's the best interpretation.
>
> That's what *I* said that you disagreed with, wasn't it?

I didn't really disagree, I just said that's not what I had in mind
originally.

In any case, I'm not sure about erasures "dropping out". I agree that
the free modifier has to attach to SA. That doesn't mean it vanishes
or that it has no meaning.

mu'o mi'e xorxes




__
Do you Yahoo!?
Yahoo! Mail - now with 250MB free storage. Learn more.
http://info.mail.yahoo.com/mail_250


posts: 14214

On Thu, Jan 13, 2005 at 04:20:06PM -0800, Jorge Llamb?as wrote:
>
> --- Robin Lee Powell wrote:
> > > > Which is more or less equivalent to the modifiers dropping
> > > > out (along with the replaced term and the sa).
> > >
> > > I think that's the best interpretation.
> >
> > That's what *I* said that you disagreed with, wasn't it?
>
> I didn't really disagree, I just said that's not what I had in
> mind originally.
>
> In any case, I'm not sure about erasures "dropping out". I agree
> that the free modifier has to attach to SA. That doesn't mean it
> vanishes or that it has no meaning.

True, but from the point of view of the parser, at least, it must
vanish with sa itself, or it's modifying something else.

-Robin


posts: 1912


> On Thu, Jan 13, 2005 at 04:20:06PM -0800, Jorge Llamb?as wrote:
> > In any case, I'm not sure about erasures "dropping out". I agree
> > that the free modifier has to attach to SA. That doesn't mean it
> > vanishes or that it has no meaning.
>
> True, but from the point of view of the parser, at least, it must
> vanish with sa itself, or it's modifying something else.

Hopefully the parser will have an option where the SA'd text
remains visible. In some cases it may be important to keep
a record of all that was said, whether or not it was later
erased.

mu'o mi'e xorxes





__
Do you Yahoo!?
Yahoo! Mail - You care about security. So do we.
http://promotions.yahoo.com/new_mail


posts: 14214

On Thu, Jan 13, 2005 at 04:34:50PM -0800, Jorge Llamb?as wrote:
>
> --- Robin Lee Powell wrote:
> > On Thu, Jan 13, 2005 at 04:20:06PM -0800, Jorge Llamb?as wrote:
> > > In any case, I'm not sure about erasures "dropping out". I
> > > agree that the free modifier has to attach to SA. That doesn't
> > > mean it vanishes or that it has no meaning.
> >
> > True, but from the point of view of the parser, at least, it
> > must vanish with sa itself, or it's modifying something else.
>
> Hopefully the parser will have an option where the SA'd text
> remains visible.

It already does. It has a *large* number of output options, in
fact.

But in the case where the SA is *not* visible, the things that bind
to SA as free and UI and so on should not be visible as well.

-Robin


posts: 14214

On Mon, Jan 10, 2005 at 10:42:51AM -0800, wikidiscuss@lojban.org wrote:
> links (be)
> linkargs (bei)

Backwards, for the record.

-Robin


posts: 14214

On Mon, Jan 10, 2005 at 10:42:51AM -0800, wikidiscuss@lojban.org wrote:
> term
> bridi-tail
> sentence
> relative-clause
> links (be)
> linkargs (bei)
> joik-ek (sumti connective)
> gihek (bridi-tail connective)

My current version does all of those except sentence, joik-ek (which
is actually going to have to be ek and joik seperately) and gihek.

Test please. http://teddyb.org/~rlpowell/hobbies/lojban/grammar/

-Robin

--
http://www.digitalkingdom.org/~rlpowell/ *** http://www.lojban.org/
Reason #237 To Learn Lojban: "Homonyms: Their Grate!"
Proud Supporter of the Singularity Institute - http://singinst.org/


posts: 14214

On Thu, Jan 13, 2005 at 05:44:08PM -0800, Robin Lee Powell wrote:
> On Mon, Jan 10, 2005 at 10:42:51AM -0800, wikidiscuss@lojban.org
> wrote:
> > term
> > bridi-tail
> > sentence
> > relative-clause
> > links (be)
> > linkargs (bei)
> > joik-ek (sumti connective)
> > gihek (bridi-tail connective)
>
> My current version does all of those except sentence, joik-ek
> (which is actually going to have to be ek and joik seperately) and
> gihek.

Other things we should probably include. Please let me know if I'm
insane or missing anytihng.

All the eks:

ek
gihek
jek
joik
interval
gek
guhek
gik

Forethought termsets:

termset

mekso:

rp-clause (which I just made up; it's fu'a plus rp-expression)
operator
operand

-Robin

--
http://www.digitalkingdom.org/~rlpowell/ *** http://www.lojban.org/
Reason #237 To Learn Lojban: "Homonyms: Their Grate!"
Proud Supporter of the Singularity Institute - http://singinst.org/


posts: 14214

On Fri, Jan 14, 2005 at 11:42:19AM -0800, Robin Lee Powell wrote:
> All the eks:
>
> ek
> gihek
> jek
> joik
> interval
> gek
> guhek
> gik

It shuold be noted that these are technically words, not structures,
and hence in some sense violate the concept of erasing by
structures.

However, I think we'd have some seriously annoyed users if
"mi broda da .e de di do du sa .e de .e di do du" didn't work.
Certainly to a user it *seems* like "ek plus some stuff" is a
structure. IMO. Any problems?

-Robin



posts: 1912


> Other things we should probably include. Please let me know if I'm
> insane or missing anytihng.
>
> All the eks:
>
> ek

As joik-ek. In other words, {ko'a e ko'e sa joi ko'e} should
give {ko'a joi ko'e}.

> gihek

Yes.

> jek

I wouldn't particularly like to include this.

> joik

Yes, as part of joik-ek.

> interval

Yes, as part of joik-ek.

> gek

It should be already included in SA term / SA bridi-tail / SA sentence.
Separating these last two may be tricky.

> guhek

{sa gu'e} should already be one case of {sa bridi-tail}.

> gik

Yes.

> Forethought termsets:
>
> termset

Forethought termsets are included as terms. I'm not sure
about {SA ce'e}.

> mekso:
>
> rp-clause (which I just made up; it's fu'a plus rp-expression)
> operator
> operand

Probably.

mu'o mi'e xorxes




__
Do you Yahoo!?
Meet the all-new My Yahoo! - Try it today!
http://my.yahoo.com




posts: 1912



> > ek
> > gihek
> > jek
> > joik
> > interval
> > gek
> > guhek
> > gik
>
> It shuold be noted that these are technically words, not structures,
> and hence in some sense violate the concept of erasing by
> structures.

They are structures actually, even if not very major ones.
ek can be as complicated as {nase.unai} for example.

But I think {joik-ek} is the structure we want SA to recognize:
any term connective should be able to replace any other.
Also, all the (ek / joik) in the grammar should be replaced by
joik-ek, and the same for (jek / joik) and joik-jek

mu'o mi'e xorxes




__
Do you Yahoo!?
Yahoo! Mail - Find what you need with new enhanced search.
http://info.mail.yahoo.com/mail_250


posts: 14214

On Fri, Jan 14, 2005 at 12:10:07PM -0800, Jorge Llamb?as wrote:
>
> --- Robin Lee Powell wrote:
> > Other things we should probably include. Please let me know if
> > I'm insane or missing anytihng.
> >
> > All the eks:
> >
> > ek
>
> As joik-ek. In other words, {ko'a e ko'e sa joi ko'e} should give
> {ko'a joi ko'e}.

OK.

> > jek
>
> I wouldn't particularly like to include this.

Whyever not?

> > joik
>
> Yes, as part of joik-ek.
>
> > interval
>
> Yes, as part of joik-ek.

Oh, didn't know interval was in there. You're right.

> > gek
>
> It should be already included in SA term / SA bridi-tail / SA
> sentence. Separating these last two may be tricky.

Can you expand on that last bit?

> > guhek
>
> {sa gu'e} should already be one case of {sa bridi-tail}.

Sort of, yeah. I suppose it'll do.

Example:

mi broda gu'e brodi gi brodo da de di sa gu'e brodo gi brodi

gives:

mi gu'e brodo gi brodi

> > Forethought termsets:
> >
> > termset
>
> Forethought termsets are included as terms.

Yes, but we can't get back to the beginning of a forethought termset
that way. We'll end up replacing the last term in it.

> I'm not sure about {SA ce'e}.

Doesn't work. I'm limited in the extent to which I care.

-Robin

--
http://www.digitalkingdom.org/~rlpowell/ *** http://www.lojban.org/
Reason #237 To Learn Lojban: "Homonyms: Their Grate!"
Proud Supporter of the Singularity Institute - http://singinst.org/


posts: 1912


> > > jek
> >
> > I wouldn't particularly like to include this.
>
> Whyever not?

We can't, in general, go back to the middle of a tanru.
I wouldn't want an exception for tanru with jeks.

> > > gek
> >
> > It should be already included in SA term / SA bridi-tail / SA
> > sentence. Separating these last two may be tricky.
>
> Can you expand on that last bit?

{ge ko'a broda gi ko'e brode} is a bridi-tail. It's not much of
a problem actually, we would use {i ge ko'a broda gi ko'e brode}
if we wanted it to count as a sentence.

> > > Forethought termsets:
> > >
> > > termset
> >
> > Forethought termsets are included as terms.
>
> Yes, but we can't get back to the beginning of a forethought termset
> that way. We'll end up replacing the last term in it.

I thought you meant SA + termset. It will replace the last term, yes.

> > I'm not sure about {SA ce'e}.
>
> Doesn't work. I'm limited in the extent to which I care.

There's also {SA pe'e} to consider. I suppose adding rules for
these two wouldn't hurt.

mu'o mi'e xorxes





__
Do you Yahoo!?
Yahoo! Mail - now with 250MB free storage. Learn more.
http://info.mail.yahoo.com/mail_250


posts: 14214

> > > > Forethought termsets:
> > > >
> > > > termset
> > >
> > > Forethought termsets are included as terms.
> >
> > Yes, but we can't get back to the beginning of a forethought
> > termset that way. We'll end up replacing the last term in it.
>
> I thought you meant SA + termset. It will replace the last term,
> yes.

Erm, yes. That's my point. I would *like* SA + nu'i li'o to
replace the last nu'i li'o.

> > > I'm not sure about {SA ce'e}.
> >
> > Doesn't work. I'm limited in the extent to which I care.
>
> There's also {SA pe'e} to consider. I suppose adding rules for
> these two wouldn't hurt.

I suppose. I am disadvantaged in not actually understanding how
they work; can you take a crack at them?

-Robin


On Monday 10 January 2005 13:42, wikidiscuss@lojban.org wrote:
> The construct to be erased need not have been completed,
> this will also work:
>
> mi viska le sa la djan
> I saw the, I mean, John.

What about {mi viska le djan. sa la djan}?

phma
--
AS d- s-: a+ c+++ p+ t f S+ e++ h r->++ n-(++)* i P- m++ M+


posts: 14214

On Sun, Jan 16, 2005 at 01:46:52PM -0500, Pierre Abbat wrote:
> On Monday 10 January 2005 13:42, wikidiscuss@lojban.org wrote:
> > The construct to be erased need not have been completed, this
> > will also work:
> >
> > mi viska le sa la djan I saw the, I mean, John.
>
> What about {mi viska le djan. sa la djan}?

Simple term replacement. You get {mi viska la djan} back in my
current parser.

-Robin

--
http://www.digitalkingdom.org/~rlpowell/ *** http://www.lojban.org/
Reason #237 To Learn Lojban: "Homonyms: Their Grate!"
Proud Supporter of the Singularity Institute - http://singinst.org/


posts: 14214

On Fri, Jan 14, 2005 at 11:42:19AM -0800, Robin Lee Powell wrote:
> Forethought termsets:
>
> termset

Done with ce'e and pe'e, but not nu'i.

> mekso:
>
> rp-clause (which I just made up; it's fu'a plus rp-expression)

Not done, actually; did mex itself instead. mex-start is the only
place that calls another -start (operand-start).

> operator
> operand

Done.

So far as I know, this completes the coding for xorsa.

One wierd thing I've seen is that {mi broda lo nu brode vau de di sa
ka brode} doesn't do what I'd like (the result is {mi broda lo nu ka
brode}. I'd like a way to handle this situation, but I haven't
thought of anything yet.

-Robin

--
http://www.digitalkingdom.org/~rlpowell/ *** http://www.lojban.org/
Reason #237 To Learn Lojban: "Homonyms: Their Grate!"
Proud Supporter of the Singularity Institute - http://singinst.org/


posts: 14214


Just for the record, here's the list of structs my parser currently
does sa on:

sentence-sa
bridi-tail-sa
pehe-sa
cehe-sa
term-sa
relative-clause-sa
linkargs-sa
links-sa
mex-sa
operator-sa
operand-sa
gihek-sa
joik-ek-sa

Oh, and we haven't done I and NIhO yet. I think those should be
done by word, and default back to the start of text if necessary.

-Robin

--
http://www.digitalkingdom.org/~rlpowell/ *** http://www.lojban.org/
Reason #237 To Learn Lojban: "Homonyms: Their Grate!"
Proud Supporter of the Singularity Institute - http://singinst.org/


posts: 1912


> Oh, and we haven't done I and NIhO yet. I think those should be
> done by word, and default back to the start of text if necessary.

The levels of sentence separation are:
I ... BO
I joik-jek
I
NIhO
NIhO NIhO
NIhO NIhO NIhO
etc.

It seems arbitrary for SA to separate the first three from the rest.
If SA is to distinguish I from NIhO it could also distinguish
{I ... BO} and {I joik-jek}, and also deal with multiple NIhO
separately.

(Do we know what something like {ni'o no'i ni'o} means, or
even {no'i no'i}?

mu'o mi'e xorxes




__
Do you Yahoo!?
Take Yahoo! Mail with you! Get it on your mobile phone.
http://mobile.yahoo.com/maildemo


posts: 1912


> One wierd thing I've seen is that {mi broda lo nu brode vau de di sa
> ka brode} doesn't do what I'd like (the result is {mi broda lo nu ka
> brode}. I'd like a way to handle this situation, but I haven't
> thought of anything yet.

I think that's because we are not used to thinking of NUs as selbri.

I suppose we could have a special rule for SA NU to replace the last
NU, but I think I prefer to treat all selbri the same way and just
learn to treat NU as just another selbri.

mu'o mi'e xorxes





__
Do you Yahoo!?
Meet the all-new My Yahoo! - Try it today!
http://my.yahoo.com




posts: 14214

On Sun, Jan 30, 2005 at 09:03:21AM -0800, Jorge Llamb?as wrote:
>
> --- Robin Lee Powell wrote:
> > Oh, and we haven't done I and NIhO yet. I think those should be
> > done by word, and default back to the start of text if
> > necessary.
>
> The levels of sentence separation are:
> I ... BO
> I joik-jek
> I
> NIhO
> NIhO NIhO
> NIhO NIhO NIhO

  • No*.


NIhO does *not* seperate sentences, or even statements, it seperates

  • paragraphs*, each of which can be more than one statement, which

can be more than one sentence each.

-Robin


posts: 1912


> On Sun, Jan 30, 2005 at 09:03:21AM -0800, Jorge Llamb?as wrote:
> >
> > The levels of sentence separation are:
> > I ... BO
> > I joik-jek
> > I
> > NIhO
> > NIhO NIhO
> > NIhO NIhO NIhO
>
> *No*.
>
> NIhO does *not* seperate sentences, or even statements, it seperates
> *paragraphs*, each of which can be more than one statement, which
> can be more than one sentence each.

Yes, and {I joik-jek} separates *statement-2's*, each of which can be
more than one statement-3, which can be more than one sentence each.

And {I} separates *statements*, each of which can be more than one
statement-1, which can be more than one sentence each.

The paragraph level gets a special name, but otherwise it is just
one more level of sentence grouping.

mu'o mi'e xorxes




__
Do you Yahoo!?
The all-new My Yahoo! - What will yours do?
http://my.yahoo.com


posts: 14214

On Mon, Jan 31, 2005 at 05:05:39AM -0800, Jorge Llamb?as wrote:
> The paragraph level gets a special name, but otherwise it is just
> one more level of sentence grouping.

OK, I'll accept that, but then the question is, what level do we sa
back to? "ni'o broda .i brode sa .i brodi" == what?

-Robin

--
http://www.digitalkingdom.org/~rlpowell/ *** http://www.lojban.org/
Reason #237 To Learn Lojban: "Homonyms: Their Grate!"
Proud Supporter of the Singularity Institute - http://singinst.org/


posts: 1912


> On Mon, Jan 31, 2005 at 05:05:39AM -0800, Jorge Llamb?as wrote:
> > The paragraph level gets a special name, but otherwise it is just
> > one more level of sentence grouping.
>
> OK, I'll accept that, but then the question is, what level do we sa
> back to? "ni'o broda .i brode sa .i brodi" == what?

I think that one can only be {ni'o broda .i brodi}.

Now, I would say {ni'o broda .i brode sa ni'o brodi}
has to be {ni'o broda ni'o brodi}, i.e. sa erases
just the last sentence.

If {sa ni'o} was to go back to the previous ni'o, then
{sa ni'o ni'o} would have to go back to the previous
{ni'o ni'o}, and {sa i joik-jek} to the previous
{i joik-jek} (or bare {i}, or ni'o), and {sa i} to the
previous bare {i} or {ni'o}, and so on.

mu'o mi'e xorxes




__
Do you Yahoo!?
The all-new My Yahoo! - What will yours do?
http://my.yahoo.com