Coding errors and omissions

[Brian MacWhinney]

Dear Sabrina,

  Comprehensive error coding is a very difficult task, so it is not
surprising to see that it has been done so infrequently.
  I like your idea of having MOR recognize 0is.  However, your example uses
0aux instead of 0is and MOR will not be happy with just 0aux, so let's go
for 0is.  Leonid is not recompiling CLAN and MOR to allow for this.  So, if
you get a version tomorrow, this will be possible.  I will also add a file
to the English lexicon called zero.cut with the obvious possible omissions.
If you find more, you would have to add more, since this will be a
non-analytic aspect of MOR at this time.  Later, we could make it more
analytic if needed.
  Regarding "me do it" that is not an error on the level of MOR.  It is a
syntactic error in case marking.  Recognizing that will take a higher level
of program, so I would just put a [*] on the main line after me.

--Brian MacWhinney


On 3/22/04 1:34 PM, "Sabrina Dunn" <sdunn+ at pitt.edu> wrote:

> Hi all,
>
> I am new to CHAT and CLAN. I have been looking through a lot of the work
> already in the CHILDES system, and I can't find consistent error or omission
> coding in any child language transcripts.  In the manual, of course, I see
> that there are several ways to code these.  In general, it is very important
> to our data that cases of missing parts of speech and missing morphological
> inflections are easy to retrieve and analyze.  Do I need to use error codes
> for every occurrence that I want to be counted, or are there CLAN analysis
> programs that recognize missing parts of speech, or just plain ungrammatical
> constructions, like a missing plural, or bad agreement?
>
>
>
> I have tried using a 0word or 0part_of_speech on the main line, but the MOR
> program ignores it.
>
> For example,
>
> *CHI:     I 0aux drawing .
>
> %mor:   pro|I v|draw-PROG .
>
>
>
> I would like to see something like this come up:
>
> %mor:   pro|I 0aux|be&1s v|draw-PROG .
>
>
>
> Should I manually do this for every line?  Will it violate the on-to-one
> mapping of morphological coding?  Does this just not need to be done,
> because there are programs that recognize it without the 0aux?
>
>
>
> Also, when an utterance reads:
>
> *CHI:     me do it.
>
> %mor:   pro|me v|do pro|it.
>
>
>
> Will a program recognize the error and track it, or do I have to manually
> code this error?
>
>
>
> Any guidance would be hugely appreciated-
>
>
>
> Thanks,
>
> Sabrina Dunn
>
> Department of Psychology
>
> University of Pittsburgh
>
>