Sign Proccessing Software

[Valerie Sutton]

SignWriting List
June 15, 2004

Dan Parvaz wrote:
> Word-completion is slightly easier when there are a fixed number of
> words/morphological variations. Simply exhausting all the varieties of
> "GIVE" would be, well, exhausting. If for purposes of word processing,
> we could agree on a fixed number of (third-person) locations, that
> might be a big help.

Right now, I think this would be a little overwhelming for the
dictionary creators. Because we are so new at creating dictionaries, I
think just letting people add signs as they learn them is probably the
best bet - Maybe in a decade we can start weeding through existing
dictionaries and improving their entries...


>
> Another thing worth considering is building probabilistic models of
> ASL text (Markov Models, say) so that there is even more predictive
> power in the typing. In other words, there could be two sets of
> constraints at play: one at the syntactic level (what words are likely
> to go here?) and one at the graphemic level (given the symbols already
> entered in this word, what is the word likely to be?). The more
> conventionalized the writing gets, the smarter we can get about this,
> and the more we'll know about SW from an information-theoretic
> viewpoint (i.e., what is the entropy of location in a signed language?
> Is is the same as that of handform?). Cool questions, IMHO.

Yes. Cool questions indeed! Give us all a decade...smile....;-)


>
> One caveat: if we guide the entry process at the word-processing
> level, it seems not only possible but likely that we would be guiding
> the evolution of SW conventionalization -- and if years of Star Trek
> have taught me anything, it's "don't mess with the timeline" :-).
> There is plenty of HCI work to suggest that people will model their
> inputs to a machine based on the feedback they receive (only natural,
> since we want to be understood with a minimum of correction). Add to
> that the fact that SW doesnt yet have a well-established tradition of
> writing outside the word processing domain (aside from a few pockets
> here and there, but nothing compared to, say, any modern European
> language), and it bears discussion whether or not this is something we
> want to do.


I agree. It would be too confining right now, because only complete
freedom will help the writing system evolve into the writing system it
should be...for us to determine constraints, before we know them, would
hurt the writing system...People need to be able to compose documents
as they choose to write them right now...Maybe in a decade, we will be
able to do more...but even in English, I am typing this email without
some program pre-guessing what the spellings would be - yuk! I would
hate to have a machine finish my words for me - Val ;-)