morphology algorithm test

[Logical Language Group <lojbab@GREBYN.COM>]

Nora has finally finished the program she wrote to test the morphology
algorithm, and will try to write up the needed changes to the algorithm
as published in JL16 this weekend.

She has done the 'obvious' checks, and it turns out that a computationally
complete test is beyond the scope of even my new hyperspeed machine - too
many possible word forms out to say a 15-phoneme string series needed to
test all of the algoirthm possibilities, even with intelligent weeding of the
test cases (which only cuts 99% of the cases out).

The best thing we can thus suggest is to have people interested, and
capable of running DOS software (its written in Turbo Pascal, and the
object is small enough to reasonably email - about 14K, even if the source
is more like 60-100K and hence a bit much to consider porting to another
language or compiler), and interested in the questions of the morphology
and/or Lojban speech recognition, let me know and we can send people
copies to bang on with their worst imagined test cases after looking at the
algorithm.

lojbab