German MOR

Brian MacWhinney macw at mac.com
Fri Dec 30 21:54:07 UTC 2005


Dear Linda,

      You can do this in MOR, but it will lead to a fairly serious  
rearrangement of virtually every arule in the language.  The result  
would be
a much less usable MOR system for German.  The basic idea underneath  
MOR is that it is supposed to be analyzing target language morphemes,  
rather than errors.  Instead of writing a rule for every possible  
error, it is better to tag errors directly In the case of spelling  
errors, the recommended treatment is simply to have

zwoelf [: zwölf]

if you want to typologize the errors or spelling types, you can have

gros [: groß] [* sß]

Here the symbol sß can be anything you want to serve as a ready tag  
for spotting error types.

In all these cases, MOR will only process the second form, but you  
retain the student's original in case you want to look at spelling  
errors or patterns.

We have also, from time to time, thought about adding a second pass  
to MOR that would take the allomorphs generate by the arules and  
submit them to a further set of across-the-board generations.  This  
would work well, for example, for the spelling changes in Romance verbs.
However, in practice, we were able to handle such phenomena well  
enough in the current framework and were not that motivated to do the  
work involved in adding this feature to MOR.

--Brian MacWhinney


On Dec 30, 2005, at 2:00 PM, Linda Forrest wrote:

> We are using German MOR to analyze samples from learners who may  
> follow different spelling conventions with regard to umlauts, sharp  
> s, etc.  Thus, we need to
>
> recognize both ‘Zwölf’ and ‘Zwoelf’ or ‘groß’  and ‘gross‘, as  
> examples. Of course, we could put both forms in the lexicon, but it  
> would be nice to geneate the second form by rule.
>
> However, as I read the Clan Manuel, once an Arule has been applied  
> to the stem, no other Arules can apply, so words with both ‘ö’ and  
> ‘ß’ would not be processed correctly. Am I missing something here,  
> or is there some other way to deal with this issue?
>
>
>
>
>
> Linda B. Forrest, Ph.D
>
> Research Director
>
> Center for Applied Second Language Studies
>
> 5290 University of Oregon
>
> Eugene, OR 97403
>
> Tel: 541-346-5699
>
> Fax: 541-346-0322
>
>
>
>
>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/chibolts/attachments/20051230/3a629fb1/attachment.html>


More information about the Chibolts mailing list