Query on Sign Documents

Wed Aug 31 01:45:39 UTC 2005

I'm confused by your differentiation between sign spelling and sign spelling sequence.

A-T-E comes before
E-A-T comes before
T-E-A 

(this is historical alphabetization)

X handshape on dominant hand with orientation 1 comes before
X handshape on dominant hand with orientation 1 with Y handshape
X handshape on dominant hand with orientation 2 comes before
X handshape on dominant hand with orientation 3

Just those four, divide out your levels to two levels.

While you cannot control "how" people create signs, they still use one sign or another, and they agree that "x handshape" in "y orientation" at "z" location is the same meaning in a given signed language. 

If you add motion, and order them, even arbitrarily, you get enough differentiation for down to almost all the signs in the Libras dictionary I have.

If you add a second hand, an orientation, and then a speed, you get all the way down most signs in the current IMWA.  

IF you add facial expression, you cover a lot more.  Even arbitrary is better than no order.

Charles

X-1 handshape with y orientation comes before 

Steve Slevinski <slevin at signpuddle.net> wrote:
Hi Charles,

I remeber reading Senor Rocha's paper a while ago. Interesting ideas. 
I'll probably read it again next year when I'm ready to tackle more 
complicated searching and matching.

I do not believe that sign spellings in a document or dictionary will be 
uniform from start position to finish position. We can't control how 
people create signs. And I think we are mixing two different ideas that 
overlap:: spelling and sequence.

Sign Spelling is symbols in space for display.

Sign Spelling Sequence is symbols in order for sorting.

The spelling and the sequence will not always have the same symbols. 
Simple sequences use only the symbols in the spelling. Detailed 
sequences use additional symbols not found in the spelling. I believe 
that Val has found with SignBank that simple sequences break down when a 
dictionary reaches about 500 signs. After that, detailed sequences are 
required to differentiate between similar signs.

So your question/opinion below concerns documents and dictionaries. 
Below I'll discuss ideals on how I think the software should work and 
options it should provide.

Signs and dictionaries...
------------------------
Spell a sign and add it to the dictionary. The sign exists in the 
dictionary, but is hard to find. You need to mark it up to make it 
easier to find. You can create either a simple or detailed sequence for 
sorting. You can add keywords, terms, or descriptions in a written 
language for searching. You can add a video for demonstration. You can 
add variants for alternate spellings of the sign. You should also be 
able to add links to synonyms.

Signs and documents...
-----------------------
Document size is a concern for some, but I'm more concerned with using a 
document than saving space. I would probably stuff all the information 
I could into a document to make it easier to use and search and link.

The signs in the documents must contain the spelling, but they may also 
contain the sequence. Signs can be linked to a dictionary sign with 
timestamp for validation. It will be easy to lookup signs that are 
linked to a dictionary. It should be easy to search for signs that are 
not linked to a dictionary. If a sign is found, the document can be 
updated. If the sign is not found, it can be added to the dictionary. 

Anyway, a few random thoughts,
-Steve

Charles Butler wrote:

> I agree, the sign-capture process is not trivial. I remember Senor 
> Rocha writing the seminal paper on the programming involved. However, 
> I have a few questions.
> 
> 1) If the sign-spelling of a sign were uniform (i.e., always start 
> with the dominant hand, etc.) and tagged as such in the creation 
> process (i.e., when a sign goes into the dictionary, whether on Sign 
> Puddle or DOS 4.5) could it not be possible, at that point, to 
> assemble a master dictionary of a given text. The problem I see, now, 
> is that a person cannot save, on SignPuddle, an unknown phrase for 
> parsing later. Example, if I see a combination of handshapes and 
> movements, in a language I don't know, and write them down, and then 
> ask "using the corpus I already have recorded" are there "any matches 
> to what I have", then by paramater A - handshape and orientation, i'd 
> get X number of matches. If there are no matches there, then I have a 
> new sign and should go through it and tag it into Sign-Spelling Order 
> for later investigation. Whether I know the meaning of a sign or not, 
> I should still be able to write it down as S-1 (Starting Position 1), 
> F-1 (! Finished Position) + movements to get from S-1 to F-1. I'd do 
> the same thing for an English word "zarf" to go between "zany" and 
> "zoo". I don't need to know a meaning to write it down and 
> alphabetize it.
> 
>
>
> */Valerie Sutton /* wrote:
>
> SignWriting List
> August 29, 2005
>
> No it is not trivial...that is for sure! But one good way to begin
> are the SignSpelling features that you are adding to SignPuddle 2.0.
> Once the person who enters the signs into SignPuddle establishes the
> sort order for each SignSpelling, then we could perhaps search on
> signs with that spelling-sort-order....someday? Val ;-)
>
> ---------------------
>
>
> On Aug 29, 2005, at 10:29 AM, Steve Slevinski wrote:
>
> > Hi Charles,
> >
> > I do not know of any true sign documents on the web at this time.
> > I believe the best way to accomplish this would use XML to store
> > the data and XLT to display the data. A while back, someone posted
> > an XLT for SWML-S 1.0. It worked in Firefox, but I think that's
> > about it.
> >
> > I believe we have two problems that need to be addressed. The
> > first problem is displaying the signs in the browsers. This will
> > probably require plug-in development. The second problem is sign
> > matching. I believe a few papers have been written on this topic.
> > It is not trivial.
> >
> > for what it's worth,
> > -Steve
> >
> > Charles Butler wrote:
> >
> >
> >> Can one yet make a sign document on the Web, and save it as a true
> >> document, so that it can be searched for signs, like a word
> >> processor? Right now, we are still limited to signs with word-to-
> >> word translation.
> >> Charles
> >>
> >> */Valerie Sutton /* wrote:
> >>
> >> SignWriting List
> >> August 27, 2005
> >>
> >> The SignMail program in SignPuddle (that let's us send email in
> >> SignWriting), can be used as a document-creating program too. For
> >> example, Nan! a from the Philippines sent the Filipino song as an
> >> email. Then I created a screen capture from the email and posted
> >> the screen-capture on the web:
> >>
> >>
> >>> http://www.signwriting.org/philippines/philippines03.html
> >>>
> >>
> >>
> >> You can also do this with the Translate program in SignPuddle.
> >> Take a screen capture once the translation is done, and that
> >> screen capture can become the document...
> >>
> >> One of many ways to use SignPuddle!
> >>
> >> Thank you, Steve, for all these features...
> >>
> >> SignPuddle
> >> http://www.SignBank.org/signpuddle
> >>
> >>
> >>
> >> Val ;-)
> >>
> >>
> >>
> >
> >
>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/sw-l/attachments/20050830/70475b88/attachment.htm>