MOR manual?

[Jennifer Ganger]

Hello,

I am wondering if there is still a MOR manual and if so, where I could find it.

Thanks,

Jenny

[Brian Macwhinney]

Dear Jenny,

    Nearly two years ago, we shifted away from using MOR to using Universal Dependencies (UD) which is going to give a very different MLU count.   I’ve posted various messages Have you been able to track any of this.  I would be happy to explain in detail.  I do have a copy of the MOR manual, but given the fact that all of the data in all of TalkBank is now in UD format, the older MOR format is not that relevant.  

—Brian

--
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 visit https://groups.google.com/d/msgid/chibolts/6f3be907-2b85-497c-90e6-0b4bc2858e9fn%40googlegroups.com.

[Jennifer Ganger]

Ah, okay. 

Is there still a %mor tier? I had already poked around this mailing group archive and the Talkbank website, and I couldn't find a beginner-friendly explanation of how one could search for things like nouns or verbs in CLAN using UD. Am I thinking about this all wrong?

We can take it off the list if everyone else already knows this!

Thanks,

Jenny

[Leonid Spektor]

Yes, there is still a %mor tier. It just has different notation. The noun and verbs in UD format simply are "noun|" and "verb|". In old MOR they are "n|" and "v|". You are right that there is no easy way to get information about UD format. The Universal Dependencies web page has all the necessary information, but it is buried deep among all other clutter. On their web page https://universaldependencies.org look for Annotation guidelines and click on that link. Next on "UD Guidelines" web page click on Morphology link. Then on "Morphology: General Principle" web page you can see what you are looking for. Particularly look at the list of universal POS tags.

I hope this helps.

Leonid.

To view this discussion visit https://groups.google.com/d/msgid/chibolts/CADPx98ZoAaRwB3EUrsfSceQg6_O-Xfc9yyK_TGKxVEFcwBUchA%40mail.gmail.com.

[Jennifer Ganger]

Thank you, that is helpful! Has all the %mor tier coding been converted to this now, so if I download a corpus, I can expect to find that coding instead of the old codes?

To view this discussion visit https://groups.google.com/d/msgid/chibolts/71E80B49-5E6E-4360-8D3D-6ACCB7E6E1A1%40andrew.cmu.edu.

[Brian Macwhinney]

Yes, everything has been converted for all 15 databases and for all languages, except a few low-resource languages that do not yet have a UD grammar.

—Brian

To view this discussion visit https://groups.google.com/d/msgid/chibolts/CADPx98bXWj76hJ1ezwf2qHo5sJysqUvEFVHJyaMnJ6%3DqoKpTrg%40mail.gmail.com.

[Brian Macwhinney]

The main page for Universal Dependencies is universaldependencies.org.  At that page the links under “Understanding UD” as the most important.

In particular the links go to

1. "UPOS tags" for the list of universal part of speech tags

2.  “Deprels” for the description of the grammatical relations.

3.  “Feats” for a summary of available features.

#1 and #2 or partially universal, #3 is decidedly not.

For each language there is extensive further docementation if you click on the language in the list and then the “language documentation” page.

These codes and concepts are very widely accepted in the NLP community and also by at least some linguists.  

— Brian MacWhinney

Teresa Heinz Professor of Cognitive Psychology, 

Language Technologies and Modern Languages, CMU 

To view this discussion visit https://groups.google.com/d/msgid/chibolts/71E80B49-5E6E-4360-8D3D-6ACCB7E6E1A1%40andrew.cmu.edu.