elidable terminators and ambiguity
vpbr...@gmail.com
I am trying to understand the rules for eliding terminators.
The CLL comments on several special cases for when terminators may or may not be elided,
but the only general principle I can find is CLL 21.2 EBNF Grammar of Lojban, point 10:
// encloses an elidable terminator, which may be omitted (without change of meaning) if no grammatical ambiguity results.
Still, grammatical ambiguity doesn't seem to me to be the deciding factor. Consider these two examples.
i mi noi le mlatu ku sisku keha vau kuho cu sipna vau
I, whom the cat is seeking, am sleeping.
The "ku" should be elidable, because when it is omitted, there is only one possible place to insert it -- no grammatical ambiguity.
But, a parser says it cannot be elided, and it does look somehow strange with the "ku" omitted.
i le mlatu poi xekri vau kuho cu sipna vau
The black cat sleeps.
The "le mlatu" needs a "ku" that can go in one of two places, before or after the relative clause.
These choices are semantically equivalent, but grammatically (syntactically) different.
You can look at the parse trees and see they are different.
Still, the parser says that eliding this "ku" is OK.
Why? Why kind of ambiguity is CLL talking about here?
mihe la bremenli nohu Vincent Broman
uakci
On Aug 26, 2016 1:47 PM, <vpbr...@gmail.com> wrote:
>
> coi ro do
>
> I am trying to understand the rules for eliding terminators.
> The CLL comments on several special cases for when terminators may or may not be elided,
> but the only general principle I can find is CLL 21.2 EBNF Grammar of Lojban, point 10:
>
> // encloses an elidable terminator, which may be omitted (without change of meaning) if no grammatical ambiguity results.
>
> Still, grammatical ambiguity doesn't seem to me to be the deciding factor.
I agree — that note is misleading. Terminators can be elided only when the meaning doesn't change.
> Consider these two examples.
>
> i mi noi le mlatu ku sisku keha vau kuho cu sipna vau
> I, whom the cat is seeking, am sleeping.
This example with only the needed stuff would be «mi noi le mlatu cu sisku tu'a ke'a cu sipna».
> The "ku" should be elidable, because when it is omitted, there is only one possible place to insert it -- no grammatical ambiguity.
No, once the «ku» is removed, we get «mlatu sisku», which is a tanru.
> But, a parser says it cannot be elided, and it does look somehow strange with the "ku" omitted.
In this case, you can also use «cu». But you can't drop both.
>
> i le mlatu poi xekri vau kuho cu sipna vau
> The black cat sleeps.
This example would be «le mlatu poi xekri cu sipna».
>
> The "le mlatu" needs a "ku" that can go in one of two places, before or after the relative clause.
Relative clauses follow the LE-clause, so «ku» is assumed automagically.
> These choices are semantically equivalent, but grammatically (syntactically) different.
> You can look at the parse trees and see they are different.
> Still, the parser says that eliding this "ku" is OK.
>
> Why? Why kind of ambiguity is CLL talking about here?
>
> mihe la bremenli nohu Vincent Broman
~ uakci
> --
> You received this message because you are subscribed to the Google Groups "lojban" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to lojban+un...@googlegroups.com.
> To post to this group, send email to loj...@googlegroups.com.
> Visit this group at https://groups.google.com/group/lojban.
> For more options, visit https://groups.google.com/d/optout.
Andrew
vpbr...@gmail.com
The reason that {ku} should be elidable (based on the rule) is that there is unambiguously only one place that the required {ku} could occupy.
The phrase {mlatu sisku} can not be a tanru because {noi},{kuho} must enclose a bridi, and {le mlatu sisku ku keha vau} is not a bridi.
That clause no verb.
uakci
Nah, it doesn't work that way. First, «mi noi le mlatu sisku cu sipna» parses successfully. It means 'Me, whom the cat sleeps' (whatever the idea of a sleeped thing is). Second, Lojban doesn't have this notion of 'if this happens, then that doesn't parse'. If your «ku» hasn't been closed, the «le» will eat more than you want it to, even at the occasional costs of not parsing.
~ uakci
--
You received this message because you are subscribed to the Google Groups "lojban" group.
To unsubscribe from this group and stop receiving emails from it, send an email to lojban+unsubscribe@googlegroups.com.
vpbr...@gmail.com
As I have read more elsewhere I've since come to the conclusion that
the "ambiguity" referred to in CLL which determines whether a terminator may be elided
refers to the ambiguity in the parsing process at the location of the missing terminator,
where the parser has only one token of lookahead to make a determination (as in YACC).
An ambiguity that requires more lookahead than that to resolve is considered still ambiguous,
and it disallows eliding the terminator in question.
This covers the mlatu sisku case anyway.
Since PEG grammars and other non-YACC parsers are widely used,
and a lot of shoe-horning has to be done to make lojban go through YACC,
I'm a little doubtful whether a grammatical rule that depends on YACC and LALR1 processing is realistic and well-chosen.
mihe la bremenli pohu laho gy Vincent Broman gy
On Thursday, September 1, 2016 at 1:00:41 PM UTC-7, uakci wrote:
Nah, it doesn't work that way. First, «mi noi le mlatu sisku cu sipna» parses successfully. It means 'Me, whom the cat sleeps' (whatever the idea of a sleeped thing is). Second, Lojban doesn't have this notion of 'if this happens, then that doesn't parse'. If your «ku» hasn't been closed, the «le» will eat more than you want it to, even at the occasional costs of not parsing.
~ uakci
On Sep 1, 2016 8:15 PM, <vpbr...@gmail.com> wrote:
The {cu} is not elidable, it is optional, but leaving it in tends to make a lot of other things elidable.--
The reason that {ku} should be elidable (based on the rule) is that there is unambiguously only one place that the required {ku} could occupy.
The phrase {mlatu sisku} can not be a tanru because {noi},{kuho} must enclose a bridi, and {le mlatu sisku ku keha vau} is not a bridi.
That clause no verb.
On Friday, August 26, 2016 at 4:02:51 PM UTC-7, Andrew wrote:The reason you can't leave out {ku} in {le mlatu ku sisku} is because it is already omitting {cu}. The longest form of this phrase would be {le mlatu ku cu sisku}. Most people pretty much always omit either {ku} or {cu}, but you need at least one between the x1 and the selbri.
You received this message because you are subscribed to the Google Groups "lojban" group.
To unsubscribe from this group and stop receiving emails from it, send an email to lojban+un...@googlegroups.com.