A question about %gra tier

Brian MacWhinney macw at cmu.edu
Mon Nov 11 19:35:24 UTC 2013


Dear Zhuo,
     Both %mor and %gra lines are generated by programs trained by state-of-the-art computational linguistic methods, rather than by human eyes.  For %mor, the error rate is at about 3-4%, which is state-of-the-art or above.  For %gra, the error rate is higher, at perhaps 8-10%, depending on how you compute it.  In general, we are actively working on improving the accuracy of %gra tagging.  
    Having said that, the problem in the specific utterance you indicated from the Valian corpus, is that the transcription method did not mark the fact that the word “eating” was repeated four times.  So, a more useful transcirption would have been:
*CHI:	eating [x 4].
If the sentence had been transcribed in that fashion, then the resulting  %gra line would have made more sense.

—Brian MacWhinney

On Nov 11, 2013, at 12:52 PM, Zhuo Chen <czcindy426 at gmail.com> wrote:

> Dear ChiBolts,
> 
> I was running a command that extracts the children's utterances in which they use a subject. The command depends on the %gra tier because it marks SUBJECT. However, I found that in some utterances, even though the children do not use the subject, %gra still marks SUBJECT. I attached the my output file. Please take a look at the last line. Could you look at it? If this is a transcribing mistake, does that mean I have to correct them by hand? 
> 
> Best,
> Zhuo
> 
> -- 
> You received this message because you are subscribed to the Google Groups "chibolts" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to chibolts+unsubscribe at googlegroups.com.
> To post to this group, send email to chibolts at googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/chibolts/45d69e07-b5b2-4e0b-8bfe-c6be6c4e0398%40googlegroups.com.
> For more options, visit https://groups.google.com/groups/opt_out.
> <01b.step_4.int.comp.notwh.notimp.SUBJECT.cex>

-- 
You received this message because you are subscribed to the Google Groups "chibolts" group.
To unsubscribe from this group and stop receiving emails from it, send an email to chibolts+unsubscribe at googlegroups.com.
To post to this group, send email to chibolts at googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/chibolts/611D73F0-65F1-41BB-8257-4380729B76EE%40cmu.edu.
For more options, visit https://groups.google.com/groups/opt_out.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/chibolts/attachments/20131111/face1035/attachment.htm>


More information about the Chibolts mailing list