<html>
<head>
<meta http-equiv="content-type" content="text/html; charset=ISO-8859-1">
</head>
<body bgcolor="#FFFFFF" text="#000000">
<font face="Times New Roman, Times, serif">I've just been discussing
similar issues with Keren Rice with regard to North Slavey (Dene),
a spoken language in Canada</font>. There is a lot of variation
from one speaker to the next, and the community has to a large
extent chosen to reflect those variations in how words are spelled.
Further, they don't want to privilege one spelling as the "right" or
"standard" spelling with the others being variants, but they want to
treat all variants on an equal basis with equal status.<br>
<br>
It strikes me that the amount of variation we see in the ASL
community (and probably other sign languages) is comparable to what
exists in Dene and many other spoken language communities I've heard
about. From a linguistic point of view, it would be important to
retain variant spellings, although it would be better if we know
where each spelling came from (whose signing it represented).<br>
<br>
There is another type of normalization, however, in which two
spellings are given that represent the pronunciation, since there
are often alternate ways of writing a given sign. This type of
variation is only variation in the writing/spelling process, not in
the pronunciation of the sign. So, I'd personally be more in favor
of normalizing that type of variation than variation that represents
actual differences in how the sign is made.<br>
<br>
Ultimately, of course, this is an editorial decision to be made by
those responsible for a corpus, so I don't want to seem to be
dictating what should be done. <br>
<pre class="moz-signature" cols="72">Albert Bickford
SIL-Mexico, Linguistic Publications
SIL International, Sign Language Global Team
<a class="moz-txt-link-abbreviated" href="mailto:albert_bickford@sil.org">albert_bickford@sil.org</a>
late May through early August: 2901 University Ave Stop 8217, Grand Forks ND 58202; 701-777-0575
mid-August through mid-May: 16131 N. Vernon Dr., Tucson AZ 85739; 520-825-6131
</pre>
<br>
On 2012/07/03 8:57 AM, Steve Slevinski wrote:
<blockquote class=" cite" id="mid_4FF2FA33_5040600_signpuddle_net"
cite="mid:4FF2FA33.5040600@signpuddle.net" type="cite">On 7/3/12
8:30 AM, Dan Parvaz wrote:
<br>
<blockquote class=" cite" id="Cite_3" type="cite">It sounds like
the user-contributed corpora could use a little
<br>
"smoothing" in the form of a normalization step (manual,
automated, or a bit of
<br>
each) that standardizes the orthography a bit for purposes of
concordance-
<br>
building, search, etc.
<br>
</blockquote>
<br>
Hi Dan,
<br>
<br>
Very true. I am almost ready to add normalization functions to
SignPuddle. They will be a combination of manual and automated.
<br>
<br>
Last year I had a breakthrough with approximate searching for
SignWriting. Given any sign, the search routines can find all
other signs that are variant spellings based on position and
order. The searching uses regular expressions and is very fast.
<br>
<br>
It will be up to the end user to decide if the signs returned from
the search routines are meaningful variant that should be left
alone or if the signs returned are meaningless variants that
should be normalized to a common spelling.
<br>
<br>
Regards,
<br>
-Steve
<br>
<br>
<br>
</blockquote>
</body>
</html>