Dublex (was: Washing-machine words (was: Futurese, Chinese,

Jeffrey:
> Dublex is an engineered language whose
> primary design goal is to develop a maximally effective minimal set of root
> words.  The goal with Dublex's roots is to identify the most productive
> roots, by analyzing a long list of compounds and seeing which roots are used
> most often.  This will provide quantifiable information on the productivity
> of specific roots (not the first such analysis, as this has been done with
> Esperanto and maybe, by now, Lojban).
Do you take into account frequency (of how many tokens of the root occur
in a text of a given size) as well as productivity (the number of
compounds a root occurs in)?

I can see how your methods allow you to discard roots that prove to be
relatively useless, but what do you do with potetially useful roots
that didn't make it into your basic inventory in the first place?

Lastly, I have the impression that in natlangs that have relatively
small inventories of roots, these roots tend to have rather fuzzy
meanings that get applied to new concepts by means of chains of
polysemy. (As a ready, though imperfect, example, see the American
Heritage list of Indoeuropean roots, available online somewhere at
bartleby.com, I think.) The result is that roots tend to suggest
(with varying degrees of vagueness) rather than determine the
meanings of words. Is Dublex like this?

--And.

Replies