Neutralizing ending in word list
Brian MacWhinney
macw at mac.com
Sat Oct 9 19:25:50 UTC 2004
Dear Diane,
Sorry about the delay in responding. I guess your goal is to treat
forms like "he'd" as if they were versions of "he". You may be right
that it will be tricky to do that using the % symbols. But, before
going in that direction, I would like to think through with you and
others, the logic of the analysis. Do you really want to say that the
pronoun is the root of a cliticized form? Isn't that going against the
idea that there are really two full words being contracted here.
Wouldn't it be better to use the +p option and treat the tilde ~ for
the clitic as
a delimiter? That was the original goal underlying the +p option and
it would seem to apply well in this case.
--Brian MacWhinney
On Oct 5, 2004, at 11:58 AM, Leach, Diane (NIH/NICHD) wrote:
>
>
> Hi folks.
>
> I have been trying to get a list of "root words" by taking my
> transcripts and neutralizing the endings on the %mor line. I was
> using the following command:
>
> freq +u +k -t* +t*CHI +t*MOT +t%mor +s"%|*" +s"%|*-%%" +s"%|*~%%"
> "*.mor.pst"
>
> This works well except when I have words in the form
> n|word-ENDING~CONTRACTION, such as n|dog-DIM~v|be&3S. In this case,
> the result is that n|dog-DIM shows up in the word list (the
> contraction was neutralized, but not the diminutive ending). I have
> tried adding +s"%|*-%%~%%" to the command line and replacing
> +s"%|*~%%" with +s"%|*-%%~%%", but neither of these seems to work. In
> the first case, nothing changes, and in the second case, it
> neutralizes the endings on the complex forms (e.g.,
> n|dog-DIM~v|be&3S), but then I still have the words with contractions
> showing up in the word list (e.g., n|dog~v|be&3S).
>
> Any thoughts about how I could fix this?
>
> Thanks!
> Diane
>
>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: text/enriched
Size: 2188 bytes
Desc: not available
URL: <http://listserv.linguistlist.org/pipermail/chibolts/attachments/20041009/355d3a94/attachment-0001.bin>
More information about the Chibolts
mailing list