re signing avatars

[Sara Morrissey]

  > what does "somewhat promising" mean?

Hmm, well roughly (and you'd need to verify this with either Philippe or
some of his papers) he's done video recognition on ASL video data that is
composed of a subset of signed sentences recorded in BostonUniversity for
linguistic reasearch by Neidle et al. 1999. There are 201 video sentences
and 104 unique words. The recognition process produces semantic glosses as
output and recognised sentences are then fed into the machine translation
system. An overview of the database and the processes can be found in the
previously mentioned paper along with results for the recognition process
alone as well as when combined with the machine translation work also
described in the paper.

Hope that helps in someway, no miracles but it does seem to be going
somewhere. But that is of course open to opinion and I'd be interested in
hearing anyone else's views on this work..

Namaste,
Sara


On 18/09/2007, Dan Parvaz <dparvaz at gmail.com> wrote:
>
> I'll go dig that up, but in the meantime, what does "somewhat promising"
> mean? Nearly all the recognition material I've seen so far has been on the
> order of < 100 isolated signs (often less than half that), and so far has
> been resistant to scaling.
>
> Cheers,
>
> -Dan.
>
> On 9/18/07, Sara Morrissey <sara.morrissey2 at mail.dcu.ie> wrote:
> >
> > >we still don't have the basic necessary and sufficient conditions for
> > sign recognition (à la the Haskins and Bell Labs research in the 1940s and
> > 1950s).
> >
> > For anyone who's interested, Philippe Dreuw at RWTH Aachen University,
> > Aachen, Germany, is working on gesture recognition and I have done some
> > collaborative work with him for translating in the direction of SLs
> > (American and Irish) to Engllish. The results from the ASL data so far have
> > been somewhat promising, although of course there is still a lot of progress
> > for it to be made usable in the mainstream. Machine translation is my area,
> > not recognition, so I'm not in a position to describe this much further but
> > you can find some more information in the paper "Hand in Hand: Automatic
> > Sign Language to English Translation" in the Proceedings of Theoretical and
> > Methodological Issues in Machine Translation (TMI-07).
> >
> > Sara
> >
> >  On 17/09/2007, Fischer Susan < susan.fischer at rit.edu> wrote:
> >
> > >  There's a group at NTT (the national phone and post office!)in Japan
> > > working on "translating" in both directions between spoken or written
> > > Japanese and what they are calling Japanese sign Language.  It's really more
> > > transliterating to or from signed Japanese.  When I observed a demonstration
> > > a few years ago, the generation of the signing avatar was pretty impressive;
> > > recognition was not, and it's a much more difficult task, not only because
> > > of the corpus problem but because we still don't have the basic necessary
> > > and sufficient conditions for sign recognition (à la the Haskins and Bell
> > > Labs research in the 1940s and 1950s).  I think some patience is in order;
> > > it took about 50 years from the time speech recognition was envisaged until
> > > the time it was accurate enough to be practically useful ( e.g.,
> > > Dragon Naturally Speaking).  Computer scientists often drastically
> > > underestimate the difficulty of determining constancy in the signal.
> > >
> > >
> > > Susan Fischer
> > > Susan.Fischer at rit.edu
> > >
> > >
> > > Center for Research on Language
> > > UCSD
> > >
> > >
> > > _______________________________________________
> > > SLLING-L mailing list
> > > SLLING-L at majordomo.valenciacc.edu
> > > http://majordomo.valenciacc.edu/mailman/listinfo/slling-l
> > >
> > >
> >
> >
> > --
> > Blessed are the flexible, for they shall not be bent out of shape.
> > _______________________________________________
> > SLLING-L mailing list
> > SLLING-L at majordomo.valenciacc.edu
> > http://majordomo.valenciacc.edu/mailman/listinfo/slling-l
> >
> >
>
> _______________________________________________
> SLLING-L mailing list
> SLLING-L at majordomo.valenciacc.edu
> http://majordomo.valenciacc.edu/mailman/listinfo/slling-l
>
>


-- 
Blessed are the flexible, for they shall not be bent out of shape.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/slling-l/attachments/20070918/0a8ca665/attachment.htm>
-------------- next part --------------
_______________________________________________
SLLING-L mailing list
SLLING-L at majordomo.valenciacc.edu
http://majordomo.valenciacc.edu/mailman/listinfo/slling-l