In the context of the present discussions, it would seem that the point of semantic categorization is heuristic, not philosophical. So, all that is needed is not complete accuracy but useful groupings (typically, similar place structures, however that may be defined). And useful for sentence construction and analysis (knowing how many places is always handy -- and what they likely mean) as well as general semantic area. With a little vocabulary adjustment, I suppose a few dozen categories would be enough, though you need to show that they actually improve learning (which will be hard, since we don't know much about learning Lojban now).
On Monday, September 1, 2014 9:56 AM, Gleki Arxokuna <gleki.is.my.name@gmail.com> wrote: