Their object to SignWriting seems to be that people 1) computers
can't animate it and 2) most people don't read it.
In terms of its suitability as a candidate for use in
an [Example Base Machine Translation] system, SignWriting lacks
the explicit linguistic detail necessary for the generation of
signs using an avatar.
This is false. You can check out the VSign project from 2004:
The 2-dimensional nature of SignWriting is easy for a human to
understand, but difficult for a computer. It is possible to animate
simple sign using only the 2-dimensional layout of symbols. For
more complicated signs, it is possible to utilized the SignSpelling
Sequence to order the action, position the symbols, and add extra
information when needed.
Annotated corpora on the other hand have the
potential to carry varying degrees of granularity of linguistic
detail, therefore bypassing the need to translate using
SignWriting and then deriving such details from the resulting
I'm not sure why they see SignWriting as an intermediate step. The
paper clearly states that documentation should be provided in a
person's native language so that they can read it in their native
language. Watching a video is not reading. Handing out a piece of
paper is not the same as requiring a computer terminal.
Another issue with SignWriting is that the majority
of signers are unfamiliar with it which lowers its appeal for
use as final output translation.
This may support the idea of including the animation in the
beginning, but it does not negate the need for written material for
people to read.