AW: Design for SignPuddle 3: parts-of-speech and morphology of sign language

Fri Dec 13 18:56:26 UTC 2013

On 12/13/13, 10:20 AM, Adam Frost wrote:
> And what exactly is being approved or rejected? The spelling of the 
> sign? Or the sign itself? How would non-standard play into it?
>
Hi Adam,

SignPuddle has always been focused on the spelling of a sign rather than 
the concept of a sign.

Similar signs are related to other signs because they are spelled 
similarly not because of connections in a database.  At the end of this 
email, I've included a detailed example of how morphology works directly 
with SignWriting text.

With SignPuddle, there is an existing security model and data 
structure.  I'm trying to stay as close as possible to this model so 
that we can continue with the current SignPuddle corpus.  We have over 
100,000 sign entries and over 1 million written signs.

Each open puddle accepts new entries from any SignPuddle user. Those 
users can edit their own entries but not other people's work. Each 
puddle can have a list of editors.  These editors can edit or delete any 
entry.

How these editors behave, how they are chosen, and their interactions 
are puddle specific.  There is no way for Valerie or I to moderate each 
individual choice.  However, we are involved in granting or revoking 
editor privileges on a per puddle basis. This give a language community 
a way to clean up their dictionary on their own.

For some editorial disputes, the right answer is to split a puddle into 
2 different puddles for 2 different groups to work separately.

I am interested in evaluating the status of each sign entry so that we 
can improve the content over time and improve the quality of the search 
results.  To help editors, I've imagined a simple workflow.

For SignPuddle 3, all of the imported content will be marked as 
"provisional".  This will be the initial quality of the ASL Dictionary: 
10,000 provisional entries.

This is not a negative, but a statement of fact.  The quality of the ASL 
Dictionary is unknown. Over time, if editors choose to work on the 
provisional entries, the numbers will change.

Imagine a year later if the ASL Dictionary reported the following numbers:
200 provisional
6,000 approved
2,000 non-standards
1,000 rejected

The search results of the approved entries would be much nicer to use 
than the search results of the rejected entries.  If the approved 
entries didn't contain what I was looking for, I could search the 
non-standard or provisional entries.

The status is overloaded for both sign quality (physical production) and 
script quality (written image).  Both of these would need to be judged 
to correctly mark an entry.

To evaluate an entry for "teacher", we would evaluate both the way it is 
written and the way it is signed.  A good writing for a good sign may be 
approved.  A good writing of an unusual sign would be non-standard.  
Likewise, an unusual writing of a good sign could be non-standard as well.

If I see a provisional entry that clearly contains an error, it should 
be marked as rejected.  If that author makes a habit of checking their 
writing, they will find one of their entries has been rejected.  If they 
investigate the reason for the rejection, they will hopefully learn 
something new and be able to fix their writing.

I can understand the personal sensibilities of not wanting to judge the 
work of others or not wanting to have my work rejected; however, for the 
SignPuddle corpus to improve there must be some workflow that helps get 
us there.

I've imagined this simple workflow and I'm open to discussion or ideas.

The quality of the SignPuddle corpus is evident with the search 
results.  Without a way to filter content, there is a lot of bad writing 
included with a lot of great writing.

Regards,
-Steve

PS - Regarding morphology with SignWriting text.

If we search for "teacher" in ASL, we will find the following signs.

These signs are examples of compound morphology because they all include 
the end morpheme of person.  With SignWriting text, it is possible to 
search for this person morpheme.  One example query string is as 
follows: QS15a40520x543S22a04520x574S15a48473x543S22a14473x574

For ASL, there are 47 signs that are found with this query.

In English, they have names like:
> audiologist
> caretaker
> champ
> champion 	Comforter
> convert
> Creator
> Creator 	criminal
> destroyer
>
If we return to the original teacher list, we can see a different 
handshape is used in one of the "teacher" signs.  A different query can 
find this non-standard entry. 
QS14c40512x529S14c48459x528S22a04517x569S22a14462x571

These are the only 2 results that describe the person morpheme in this way.

If we modify the query string for the same handshapes, but in other palm 
facings and other rotations, we get more results. 
QS14cuu512x529S14cuu459x528S22a04517x569S22a14462x571

Now, all of this information was entered by SignPuddle users and 
searchable within the SignWriting Text.  It is a bit of a jungle, but a 
very beautiful and interesting place.

________________________________________________

SIGNWRITING LIST INFORMATION

Valerie Sutton
SignWriting List moderator
sutton at signwriting.org

Post Messages to the SignWriting List:
sw-l at listserv.valenciacollege.edu

SignWriting List Archives & Home Page
http://www.signwriting.org/forums/swlist

Join, Leave or Change How You Receive SW List Messages
http://listserv.valenciacollege.edu/cgi-bin/wa?SUBED1=SW-L&A=1
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/sw-l/attachments/20131213/c72b7242/attachment.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: becbifhc.png
Type: image/png
Size: 30686 bytes
Desc: not available
URL: <http://listserv.linguistlist.org/pipermail/sw-l/attachments/20131213/c72b7242/attachment.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: iaiafgbd.png
Type: image/png
Size: 4119 bytes
Desc: not available
URL: <http://listserv.linguistlist.org/pipermail/sw-l/attachments/20131213/c72b7242/attachment-0001.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: hcjiccge.png
Type: image/png
Size: 25186 bytes
Desc: not available
URL: <http://listserv.linguistlist.org/pipermail/sw-l/attachments/20131213/c72b7242/attachment-0002.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: fgieeaci.png
Type: image/png
Size: 4008 bytes
Desc: not available
URL: <http://listserv.linguistlist.org/pipermail/sw-l/attachments/20131213/c72b7242/attachment-0003.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: jahbhidb.png
Type: image/png
Size: 8946 bytes
Desc: not available
URL: <http://listserv.linguistlist.org/pipermail/sw-l/attachments/20131213/c72b7242/attachment-0004.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ffgghfcd.png
Type: image/png
Size: 30095 bytes
Desc: not available
URL: <http://listserv.linguistlist.org/pipermail/sw-l/attachments/20131213/c72b7242/attachment-0005.png>