[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Lojban - why the gismu-making algorithm
>From: jrk@SYS.UEA.AC.UK (Richard Kennaway)
>Subject: Re: CONLANG: comment on Lojban
>
>John Cowan described the method by which Lojban vocabulary was compiled.
>
>This raises a question which I have wondered about ever since Loglan
>(whose vocabulary was compiled in essentially the same way). What was
>the purpose of using the chosen algorithm, and does the algorithm
>satisfy that purpose?
I'll suggest that people with access to JCB's Loglan 1 read for
themselves, and someone who has more time can perhaps excerpt actual
quotes. In 3rd edition L1, read section 1.3, and Notes 7 and 9 of
Chapter 6. My 4th edition copies have wandered off, so I am not sure if
those section and note numbers apply to the newer edition.
Briefly, JCB cites cultural neutrality for purposes of linguistic
experimentation, and maximizing the AVERAGE recognizability over the
potential speaker population for ease of learning. JCB's approach is
>A< way to satisfy his goal - whether it is the best way or whether it
is even useful at all, is subject to dispute.
>"Maximising recognizability" was the explicitly stated purpose of the
>original Loglan algorithm, and the score for each word was called a
>measure of recognizability. The meaning of most Loglan and Lojban words
>cannot be even approximately guessed just by seeing them, but the
>recognizability is supposed to aid in vocabulary learning.
JCB's intent was that a speaker attempting to learn the words using
recognizability as an aid would be explicitly told the etymological
basis, as a first order memory hook. For this reason, LogFlash data
files use the English etymological root as the keyword (even if it is
not definitional) or put the root in as a "clue word" if the keyword is
NOT the etytmological root.
You are not intended to have to guess the roots blindly from the Lojban
without ever having been told them.
>Has any research been done to determine whether it does? E.g. one might
>compare vocabulary memorization performance for Lojban-English word
>lists with similar tasks, e.g. the same word list with a random mapping
>of L to E, a word list with randomly generated Lojban words, and the
>same for a natural language.
JCB claims to have tested various scoring algorithms back in the 50s, on
some of his college classes, but never wrote a paper or otherwise
presented his experimental data for analysis by others. Of course, he
was testing on naive native-English speakers only. He wrote a little
about this in a Lognet within the last few years - the only place where
he has discussed the procedure.
A few years ago we added instrumentation to the PC LogFlash version to
allow for some testing of the algorithm. I've called for guinea pigs,
but no one has ever to my knowledge gone through LogFlash with the
instrumented version enough to give us data. I've had volunteers,
including some from the last time this question was asked on Conlang,
but we did not have LogFlash up on the net at that time. It is now
available on our ftp site, and anyone who wants to serve as guinea pig
for the recognition algorithm should contact me, preferably before
starting, so I can maximize the chances of getting useful data from your
efforts.
If I get data from even one person, it will provide some clues as to the
efficacy of the algorithm, though a through test would require speakers
of several native languages to use the program, each woth appropriately
designed gismu lists. Jorge Llambias has been slowly working on a
Spanish language gismu list, which could eventually serve as the basis
for testing the algorithm on Spanish native speakers.
lojbab
----
lojbab lojbab@access.digex.net
Bob LeChevalier, President, The Logical Language Group, Inc.
2904 Beau Lane, Fairfax VA 22031-1303 USA 703-385-0273
For the artificial language Loglan/Lojban, see powered.cs.yale.edu /pub/lojban
or see Lojban WWW Server: href="http://xiron.pc.helsinki.fi/lojban/";