>  John Clifford gave a talk on conlangs of this type to the first Language
> Creation Conference; I think the video is online.

i watched this video a while ago. i think the languages he mentioned were
toki pona, aUI, esperanto and lojban of course.

especially toki pona and aUi seem to have been constructed with a set of
primitives that is as small as possible. but i suppose the dictionaries
which are to be written around this will either get full of ambiguities
(which for toki pona surely was intended so) or vast and hard to manage, and
the compounds will get long and/or full of metaphors (aUI: "nose" is kEmOz

the great question to me is how to define the set of primitives without
simply translate some words which seem useful to me (some numbers, colors,

I want to do something like this, but I am not good at defining words with
> her primitives.
> I hope to build a language from her primitives, from the bottom up, rather
> than top-down, the way natlang definitions are done.

but that's the point! it should be possible to take any natlang expressions
and try to 'decompose' them into primitives through definition, so that they
are embedded in whole sentences. in this way some semantic relations between
the primitives should be established as a byproduct. any categories or
classes can be applied afterwards to finally construct a morphosyntax around

is there perhaps someone who has already tried this approach? i have tried a
little for myself, but i came to the point, that it would take too much time
and effort to run such a project on one's own.