[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [lojban] New to lojban, any suggestions?
At 11:07 PM 8/12/01 -0700, Robin Lee Powell wrote:
> I want to make sure to note that people who use LogFlash are helping the
> research aspect of the project.
Uhh, really? No offense intended, but perhaps then it should exist in
a form that can actually be used by most people.
Meaning?
> LogFlash is instrumented, and anyone like Arnt who makes it through
> all the words is urged to send me the files specific to their name.
> Info on the learner's education and language background (especially
> with regard to the 6 source languages for Lojban gismu, is also
> useful.
<sigh> I can't believe I'm saying this.
If you explain in more detail why Logflash is important, and I agree
with the value as presented in said explanation, I'll give serious
consideration to re-writing it.
Rewriting it? Why?
LogFlash is instrumented to log the user's sessions, and also keeps
statistics on how many hits and errors that the user has in learning each
word. This is the first cut for research as to whether the recognition
scores used to make Lojban words actually mean anything in terms of
learnability. If we get any correlation with a small self-selected and
haphazardly learning set of users, then we likely would want to repeat the
experiment more systematically with new users (and probably modify the
instrumentation based on what we found in the first pass).
Such a second-level research effort would be specific and focussed in the
way no other Loglan/Lojban research has been, could be described in terms
that do not limit applicability to learning only Lojban, and hence could be
eligible for grant funding from some organization that funds academic
research in second language learning. This of course would build the
Loglan project's credibility in the scientific research arena, credibility
that is hard for any artificial language research to gain.
Is the research value 'tainted' if the user has done other, seperate
memorizing of the words?
Such a taint is real, but at this point I suspect that tainted data is
better than no data. Words that someone has learned before using LogFlash
will "go right to the top" with no errors, and this could be filtered out,
leaving us the rest of the data to analyze.
A secondary goal is to statistically demonstrate the effectiveness of the
LogFlash algorithm. While you and others have at times thought poorly of
it for varying reasons, LogFlash is an automation of JCB's flashcard
algorithm, which he apparently devised after conducting his own serious
research into flashcard methods to find what worked best. LogFlash is thus
(presumably) tuned for optimal learning.
BTW, it IS possible to use LogFlash with a subset of the words, by running
a weeded gismu list through a little program that builds the index that the
program uses. Not sure how good this would be for the research purposes
(if only because analyzing the data would need a smarter program that could
match the instrumentation data to actual words rather than to line number
in the gismu list), but it does avoid one of your criticisms.
lojbab
--
lojbab lojbab@lojban.org
Bob LeChevalier, President, The Logical Language Group, Inc.
2904 Beau Lane, Fairfax VA 22031-1303 USA 703-385-0273
Artificial language Loglan/Lojban: http://www.lojban.org