Problem with using EVAL and C-NNLA commands

[Jorge Ricardo Hidalgo Chagoya]

Greetings!

I am currently facing problems with extracting specific variables from my data. Importantly, my data is in Spanish. I have previously used the batchalign script to have accurate %mor and %gra tiers tags, since the tagging using CLAN´s command of MOR was not as accurate. But now that I try to use the EVAL and C-NNLA commands with the resulting chat files, several fields appear in blank (including number of verbs and nouns). Beforehand, I ran the CHECK command and corrected all detected errors. 

I realise that the parts of speech tags are different from one method to another. My best guess is that EVAL and C-NNLA are specifically designed to track specific strings in the %mor tier (for verbs: the "v" obtained by the MOR command instead of the tag "verb" obtained via batchalign). 

My main concern is not being able to use the batchalign files. Would you know if there is a way to solve this issue? Perhaps a method to modify the %mor tags so they fit the EVAL and C-NNLA requirements?

Thank you for your attention!

Jorge Hidalgo. 

[Leonid Spektor]

Hi,

Unfortunately, I don't have a good news for you. EVAL and C-NNLA were designed for %mor line created with MOR command and for English language. Both EVAL and C-NNLA track very specific tags that do not translate directly to batchalign notation. At this time the only person who can look at this and come up with translation between MOR and batchalign is on vacation until 12th of March. We are in the process of make all CLAN command able to work with batchalign notation. EVAL and C-NNLA have not been changed yet.

Leonid.

--
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+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/chibolts/b7a6dd92-2005-4479-ba94-50a2e1ae0372n%40googlegroups.com.

[Brian Macwhinney]

Jorge,
Thanks for your message. I’m glad that batchalign is working okay for you in terms of taggng.

Regarding C-NNLA, the origianl NNLA program from Northwestern was only designed and tested fo work with English. Developing a similar instrument for other languages would be a project for each language.

EVAL was also never operative for Spanish. However, making it work for Spanish would not be fundamentally difficult. It will just take some time. Having some Spanish aphasia data n AphasiaBank would be a great moiivating force getting us to make EVAL working for Spanish.

— Brian MacWhinney
Teresa Heinz Professor of Cognitive Psychology,
Language Technologies and Modern Languages, CMU

[Jorge Ricardo Hidalgo Chagoya]

Leonid and Brian,

Thank you for your fast responses!  This are indeed terrible news. I knew CNNLA was not necessarily fit for Spanish as it is stated in the manual, but I thought my problems with EVAL were derived from defects in the tagging. I am currently trying to extract some variables just using the FREQ command. 

It would be great to have more aphasia data in AphasiaBank indeed. Unfortunately, the data I am working with is not my own and I still don't know how they plan to treat the data once the collection is finished. 

Thank you again for your replies. It is really admirable the effort you put in the continuous development of this project. 

Sincerely, 

Jorge Hidalgo.