[lojban] Learning vocabulary

[Chris Capel <lojban-out@lojban.org>]

Hi everyone,

I'm interested in learning Lojban. However, I find the age, the
platform, and even the approach of the current LogFlash program to be
rather unappealing. There has been a lot of effort put into solving
memorization in general, and I think it would be better to take
advantage of that effort by using a program such as SuperMemo or
MemAid (free software, multi-platform) and simply provide the
appropriate items as raw data for these programs.

I would be willing to work on providing a list of items, but I'd like
the input of the group before I start. There are several
considerations. First, it's necessary to understand the general
approach of Supermemo-type programs to learning. In these programs,
each individual piece of knowledge is completely independent of all
others. Each time an item is reviewed, it is rescheduled for review at
a later date. The interval is based solely on how difficult the item
is for the user (how well they've scored on that item in the past) and
what the previous scheduling intervals were. Using this information,
the program tries to estimate the current forgetting curve (see
http://memaid.sourceforge.net/docs/repetitions.html) and schedule the
item optimally. New items are introduced gradually, at a pace
determined by the user.

Next, each item in these programs consists of two parts--a "question"
and an "answer", which can be any text. The question is shown to the
user, the user is prompted to confirm, the answer is shown, and the
user grades him- or herself. If the user wishes, they can provide
themselves with additional forms of sensory reinforcement, such as
repeating the answer out loud, or typing it on the keytops.

Fitting knowledge of lojban into the question/answer format shouldn't
be too difficult, but it would be quite different than items in
LogFlash. First, because of the approach of the method, it would be
best to separate each brivla (and perhaps selbri needing separate
definition) into many different items. One way to do this is to have
one item for the overall "sense" of the selbri (it would include the
place order, but the user wouldn't grade themselves as hard on it),
and one item for each place of the selbri. For place items, the
question would be composed of the selbri along with the FA cmavo
indicating the place in question. This has the additional advantage of
directly teaching the correct FA word for that place for that selbri,
instead of knowing the place's order in the selbri and having to also
think up the FA cmavo for that place, as would happen when you simply
memorize the places in the "x1 goes to x2 via x3" form.

Each item would probably need to be included twice, once with the
question and answer reversed. This could be made less monotonous by
staggering the introduction of the different items pertaining to a
word in the new word list, so that the lojban/english item is
introduced first, followed by place items, followed by english/lojban,
followed by rafsi and reversed place. So it could take weeks to
introduce all the items for a single word.

Cmavo may be more difficult to formulate questions for. For one, they
often have no useful gloss. For another, they often have difficult
definitions, and true understanding only comes with examples and
practice, and a knowledge of the overall organization of the cmavo
groups. I think introductions to the cmavo are best left to other
pedagogical tools, but a good way to practice them would be to have a
number of examples (one per item) in English and their lojban
equivalents for the question and answer. Again, these could be
reversed for additional practice.

Rafsi could be learned by simply having the gismu as the question and
its various rafsi as the answer.

Are there any other language elements that would be suited to a memory
program that I'm missing?

Next, I think it's important to allow people to learn the vocabulary
bits at a time. For that, it's necessary to have some sort of taxonomy
and provide lists of words in each category. There could also be a
list (or two--level 1 and 2) including a basic core of all of the most
frequently used words, and from there other groups would build. There
could be a group with computer science words, and a group with
practical living words, and a group with "academic" words (the kind
used in philosophy and other non-fictional exposition), and so on.
There could be an 'eclectic' group. Of course, there's no reason not
to include an item in multiple groups. A user could simply ignore
those that they have already learned when adding new groups.

Finally, I think it's important that brivla/selbri lists include lujvo
as well as gismo, from the beginning. They can be told apart by
spelling, and a lojban word list ordered by frequency of use would
include hundreds or thousands of lujvo before it included 90% of the
gismu (I'm guessing), so having all the gismu be learnt first seems
really silly to me. Of course, this assumes that there are enough
established lujvo to be included in such a list. For completeness (and
for the rafsi knowledge) there could be one vocabulary group including
all the miscellaneous gismu, with rafsi, not in the basic groups,
though they might be duplicated in the advanced groups.

The core basic group is the most essential. Mastery of it would give a
person with some grammar knowledge the ability to hold basic
conversations in lojban, and to read text with the help of a
dictionary. Thereafter they would be free to continue their education
through other methods, with much less resistance than they would have
had before. It might also, if translated, give people the ability to
continue to learn lojban in the absence of many dictionary definitions
or tutorials in their native language, allowing them to "bootstrap"
into a knowledge of lojban with relatively little translation effort
required to make it possible. (This aspect is also somewhat important
to me, personally.)

I'm willing to do as much as I can in my spare time to see this
happen. However, in interests of conserving resources, it might be
wise to somehow make this part of the jbovlaste project. Some items
may be able to be generated automatically (or mostly so) from
jbovlaste data, and many other items (probably most) will be
straightforward rearrangements of parts of the word's definition.

I can also provide programming if existing memorization programs fail
to meet the needs of lojban pedagogy.

As an excellent learning program, including tutorials, references, and
vocabulary tools, is essential for the growth of lojban, I think this
project, if it doesn't have any flaws I don't see immediately, will be
worthwhile.

I think a first step would be to identify a list of words and cmavo
examples as candidates for inclusion in a basic-level word list. I
could go through the gismu myself, but on cmavo and lujvo, I'm afraid
I wouldn't be much use.

Chris Capel
-- 
"What is it like to be a bat? What is it like to bat a bee? What is it
like to be a bee being batted? What is it like to be a batted bee?"
-- The Mind's I (Hofstadter, Dennet)