Following NoDaLiDa, I've been thinking about how to improve dependencies 
(and relations) to be done at the reading level. But, this brings some 
complications with it, and I'd like some feedback from the group.

*Cohort-Cohort*
Currently, dependency is from one cohort to parent cohort:
"<bears>"
  "bear" N PL #1->0
  "bear" V #1->0
"<dogs>"
  "dog" N PL #2->1
  "dogs" V #2->1

*Compatible upgrade: Reading-Cohort*
This could be improved to allow dependencies from each reading to a parent 
cohort, without changing the stream format. E.g.:
"<bears>"
  "bear" N PL #1->0
  "bear" V #1->2
"<dogs>"
  "dog" N PL #2->0
  "dogs" V #2->1

*Incompatible: Reading-Reading*
But a much more powerful implementation would be to allow dependencies from 
each reading to any other reading group. Keyword: group. And that's where 
the complication comes in.

Because of the way SetParent and context tests work, they find matching 
readings, plural. If you SetParent (A) To (-1 B), the B may match more than 
one reading, so each reading matching A will get all matching B readings as 
parents. Internally, this shouldn't be a problem, and is how I'd implement 
Reading-Cohort anyway, so that within a single grammar this power is 
available.

But when writing the output, how do I represent multiple parent readings? 
Clearly, it has to be a comma-delimited list of targets, but those can be 
done two ways. First way is for each reading to have a sentence unique 
number:
"<bears>"
  "bear" N PL #1->0
  "bear" V #2->3,4
"<dogs>"
  "dog" N PL #3->1
  "dogs" V #4->1,2

Second way is for each part to include the cohort number, something like:
"<bears>"
  "bear" N PL #1:1->0
  "bear" V #1:2->2:1,2:2
"<dogs>"
  "dog" N PL #2:1->1:1
  "dogs" V #2:2->1:1,1:2

...which, quite frankly, is hideous.

But the whole Reading-Reading output format would be behind a --cmd-flag 
because it can't be compatible. Reading-Cohort output could be done and 
nobody would notice in daily use.

Any comments, ideas, etc?

-- Tino Didriksen

-- 
You received this message because you are subscribed to the Google Groups 
"Constraint Grammar" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To post to this group, send email to [email protected].
Visit this group at http://groups.google.com/group/constraint-grammar.
For more options, visit https://groups.google.com/d/optout.

Reply via email to