Re: [lojban] camxes's reaction to some fu'ivla

[Jorge Llambías <jjllambias@gmail.com>]

On Sat, Jan 23, 2010 at 12:42 AM, Pierre Abbat <phma@phma.optus.nu> wrote:
> I installed camxes (the compiled jar) and ran it on some words. I gave it all
> the months of the Roman calendar (ianvari, frebuari, martio, prilio, madjio,
> djunio, djulio, avgusto, septembero, oktobero, novmbero, decmbero; kuintili
> (=djulio), sektili (=avgusto), mercedonio (noi setca ke'a lo frebuari)), and
> it accepted them all. Vlatai rejects zo martio bi'o zo djulio.

Right. camxes accepts the CS syllable onset (where S stands for the
semivowels i, u), although there has never been an official decision
one way or the other about this. Why does vlatai reject them? I
thought vlatai was generally more permissive than camxes about these
things.

> "soiombo" (a Mongolian symbol) and "peuence" (the Chilean people of the
> monkey-puzzle tree) it rejects.

It should read them as "so iombo", "pe uence".

Basically, what camxes does is treat the semivowels i/u as onsets and
thus any word that starts with them doesn't require a ".".

> These would be valid if the rule were "The
> second consonant must be next to another consonant when y'y and ybu are
> ignored", but with the current rule, "There must be a consonant cluster in
> the first five letters, ignoring y'y and ybu", they're not. However, it
> considers "soiombo" to break into "so iombo". This I think is incorrect. A
> word beginning with a vowel must be preceded by a pause, else you can't
> tell "so iombo" from "soi ombo". Both programs accept "ricrpeuence".

"soi.ombo" requires a glottal stop. "soiombo" is unambiguously "so iombo".

The rule about "the first five letters" is a bit nonsensical from a
phonological point of view. There is no sensible reason to reject
"ba'auski" if ".a'auski" is accepted.

The camxes rule basically is: if it consists of Lojban non-y syllables
with penultimate stress, it ends in an open syllable, and the initial
part doesn't fall off as a cmavo or slinku'i consonant, then it's a
valid fu'ivla. (Also, the first syllable can't be a consonantal
syllable.)

> camxes rejects "aierne". Its attempt at a parse is simply "a" (".a .ierne" is
> not the start of anything grammatical, as "a" is expecting a sumti to
> follow), even though ".ai .erne" is grammatical.

It's grammatical for example in "zo .a ierne", or "lo'u .a ierne le'u".

In any case, the lexer doesn't care about syntax, it just breaks a
stream of phonemes into words. Whether those words end up forming a
grammatical utterance or not is not a concern for the lexer.

> camxes accepts "amliau" and not "mliau";

Yes, because "am" and "liau" are valid syllables. camxes doesn't like
CCS as syllable onset.

> jbofi'e accepts "mliau" as a word and
> splits "amliau" into a misparse. Whichever one is valid means "meow".

(Although the phonological issues that still have to be officially
settled are relatively few and marginal, the BPFK should make a
decision about them at some point, it probably doesn't look very good
for Lojban that we still haven't made up our mind about them.)

mu'o mi'e xorxes