CMR normalization issue for relational nouns

[nschneid]

Re: #57

https://github.com/UniversalConceptualCognitiveAnnotation/UCCA_English-EWT/blob/f30dc0dbc9196423336dac8ea0b4f1e3e89958f3/194313-0004.svg

before normalization: https://github.com/UniversalConceptualCognitiveAnnotation/UCCA_English-EWT/blob/2083a4f00d14244390f9df4964410a6d419946c9/194313-0004.svg

I think the CMR-normalized analysis should be: [[My_A wife_S+A]_C and_N [(My)_A kids_S+A]_C]_A

[danielhers]

Thanks! I did some eyeballing of results but it's always good to have more checks. So you're saying that in case the main relation is also a participant, the normalized version should not be changed to Parallel Scenes + Linker, but rather stay Centers + Connector, right?

[nschneid]

I think so. But I'd want to see more examples to know whether that's always the case.

[danielhers]

@omriabnd what's your opinion?

[omriabnd]

I don't know. In some cases it could be ambiguous. I don't think they should have marked it as CMR to begin with. How many such cases are there in the corpus?

[danielhers]

Just 4 out of 174:

$ git grep Coordinated_Main_Rel 845017c8d |wc -l
174
$ git grep -A2 Coordinated_Main_Rel 845017c8d | grep 'tag="A"' | wc -l
4

[danielhers]

So should I change the handling in normalization.py or leave it this way?

[omriabnd]

I would leave it this way, and update the guidelines that cases like this should not be annotated with CMR.

On Fri, Mar 22, 2019 at 4:58 PM Daniel Hershcovich notifications@github.com wrote:

So should I change the handling in normalization.py or leave it this way?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/danielhers/ucca/issues/62#issuecomment-475653447, or mute the thread https://github.com/notifications/unsubscribe-auth/AIG86-LumzutIn6CpjJQ6vUsReG0DWtKks5vZO-JgaJpZM4cDpgE .

[nschneid]

Can it be flagged as an error in normalization/validation code?