On Tuesday 04 February 2003 15:48, Robert LeChevalier wrote:
> At 06:48 AM 2/4/03 -0500, Pierre wrote:
> >How would you lex the following?:
>
> My best guess, and I'm not using any algorithm per se.
>
> >/la xidEkel. rIrxe fa'o tIgri se li ni stIka lI te/
>
> # and abort processing at this point.
valfendi currently outputs what you say (plus type-of-word annotations).
If it
had fa'o-detection, it would lex it as /la xidEkel rIrxe fa'o tIgri
selinistIkalIte/ with everything from /tIgri/ on resolved as foreign text.
But the words after /fa'o/ are in fact /tIgris elinistI kalIte/, and if there
were a Greek parser which saw an instruction to call the Lojban parser and
then resume when the Lojban was finished, it would lose those three words,
since the Lojban lexer cannot detect {fa'o} until it has read the end of the
piece. That is why I think that {fa'o} should must be always followed by a
pause: not because the algorithm can't lex it otherwise (it does, as well as
lex some other illegally-pronounced phrases such as /kybuladjAn/) but because
not putting a pause forces the algorithm to process text after the {fa'o}
before it can detect the {fa'o}.