On Sun, Jul 20, 2008 at 4:54 AM, Mateusz Grotek
<unoduetre@poczta.onet.pl> wrote:
Hello.
I'm new to lojban, recently started reading CLL, but have some questions
about brivla recognition from speech stream.
What is exact algorithm for doing it? I tried to create one, but it looks
like i have to count letters before stress, what i don't wanna do. Is it
really needed? (Because of something what is called "tosmabru failure" in
book). And point 5b) in draft look for me somehow wrong, but maybe it's my
fault. Could you explain it to me please?
You can find the (an) algorithm here:
< http://www.lojban.org/tiki/tiki-index.php?page=BPFK%20Section%3A%20PEG%20Morphology%20Algorithm>
Basically it works as follows:
(1) If the speech stream that you are considering (from the start up
to the first pause)
contains some non-Lojban phoneme or an impermissibe cluster then it is
a non-Lojban
word.
(2) Otherwise, if it ends with a consonant, it is a cmevla.
(3) Otherwise, if it starts with something that could be a cmavo, and
what remains
is a possible word or words, that first part is indeed a cmavo.
(That's for example
what happens with "tosmabru", "to" is a cmavo because "smabru" is a possible
word.) But there is one exception here: if it starts with CVCy and it
is a lujvo, then
CV cannot be a cmavo (so "tosymabru" is not "to sy mabru").
(4) Otherwise, unless it is a "slinku'i", it is a brivla. A "slinku'i"
consists of a
consonant followed by a string of rafsi. A slinku'i is also a non-Lojban word.
Unfortunately, it is not possible to characterize brivla without
recourse to rafsi.
Both the slinku'i rule, and the tosymabru rule make use of rafsi strings.
mu'o mi'e xorxes