search

UniversalConceptualCognitiveAnnotation / docs

UCCA Documentation
https://universalconceptualcognitiveannotation.github.io/
10 stars 1 forks source link

"the judge": scene-evoking? #25

Open nschneid opened 5 years ago

nschneid commented 5 years ago

A student-authored example that we felt fairly confident about, except for one issue: p. 18 of the guidelines mentions that "driver" is a scene-evoking noun in "taxi driver". So should all uses of "judge" (the profession) be scene-evoking? Should it be P+A, because "judge" denotes both the scene and one of its participants? Should there be an implicit second participant, that which is judged?

img_20180913_142052

dotdv commented 5 years ago

We discussed this a few days ago with one of the annotators: we said the profession should be annotated just P (without A or IMP A). so for example: [[Judge_P]_E Harris_C]_A ruled in John's favor

nschneid commented 5 years ago

So what are the criteria for annotating P+A or S+A on a unit?

FrameNet has a concept of "denoted frame element"—basically if a noun both evokes a scene and denotes one of its roles. I figured that's what P/S+A was for.

omriabnd commented 5 years ago

Resolution: judge should be P+A, not if it's a title. Cite FrameNet in a footnote.

nschneid commented 5 years ago

What about "the handwritten letter to John"? Is "letter" scene-evoking because it implies sending a message to John?

nschneid commented 5 years ago

One option: the_E [handwritten_P (letter)_A]_E letter_C [(IMP)_P [to_R John_C]_A]_E Compare "the handwritten letter sent to John"

dotdv commented 5 years ago

One of the annotators asked me: [[[the_E judge_C]_P+A]_A listened to the witness. or [[the]_E [judge_P+A]_C]_A listened to the witness.

I saw the latter being used on several occasions also by annotators, but I think that the first option is more consistent with other instructions we give, namely that Cs can't be scenes and that determiners should be included inside the P/S (articles and demonstratives only I think). So in a more complex example: [[TheE]{P+A-} tired_D [judgeC]{-P+A}]_A listened to the witness

This will make it similar to how we generally tend to treat these types of determiners in noun scenes : We saw_P [[aE]{P-} great_D [showC]{-P}]_A *the decision to include the determiner within the P/S even when there's an adjective (D in this case) in between is something we verbally agreed upon with the Eng annotators, but haven't yet clarified in the guidelines.

omriabnd commented 5 years ago

I agree with you. The first option is better.

On Wed, Oct 3, 2018 at 10:15 AM dotdv notifications@github.com wrote:

One of the annotators asked me: [[[the_E judge_C]_P+A]_A listened to the witness. or [[the]_E [judge_P+A]_C]_A listened to the witness.

I saw the latter being used on several occasions also by annotators, but I think that the first option is more consistent with other instructions we give, namely that Cs can't be scenes and that determiners should be included inside the P/S (articles and demonstratives only I think). So in a more complex example: [[The_E]{P+A-} tired_D [judge_C]{-P+A}]_A listened to the witness

This will make it similar to how we generally tend to treat determiners in noun scenes : We saw_P [[a_E]{P-} great_D [show_C]{-P}]_A This decision of still including the determiner within the P/S even when there's an adjective (D in this case) in between is something we verbally agreed upon with the Eng annotators, but haven't yet clarified in the guidelines.

— You are receiving this because you were assigned. Reply to this email directly, view it on GitHub https://github.com/UniversalConceptualCognitiveAnnotation/docs/issues/25#issuecomment-426535204, or mute the thread https://github.com/notifications/unsubscribe-auth/AIG86yzkYfiXffl7PvhqdB6vSUzbV59Tks5uhGQAgaJpZM4WoF_G .

dotdv commented 5 years ago

There is a problem though. How should we deal with cases like: "[The tall judge]A was able to reach for the book on the top shelf" Since 'tall' doesn't modify the process of judging, marking it [[the]{P+A-} tall_D judge{-P+A}] seems problematic, no? or do you think it somehow acceptable? If indeed we see it as a problem, then it seems that in such cases we will need to first start with C and Es, but then is there a way to avoid C scenes?: [The_E [tallS (judge)_A]_E [judge\{P+A}]_C]_A was able to reach for the book on the top shelf

Maybe a possible alternative is to internally mark the Remote as P: "the_E [tall_S (judge_P)_A ]_E [judge]_C" But this would only be optional if it were possible to internally annotate Remotes..

nschneid commented 5 years ago

Good point. Let's make it even more interesting and add a modifier that should go in the scene evoked by "judge".

Suppose we're talking about a dog show, and thus refer to a "dog judge" (somebody who judges the dogs in the competition).

omriabnd commented 5 years ago

I think we should allow C Scenes, but instruct annotators to use them very sparingly. For instance,

experienced_D taxi_A driver_P+A

and not:

experienced_E [taxi driver]_C

On Thu, Oct 4, 2018 at 5:48 PM Nathan Schneider notifications@github.com wrote:

Good point. Let's make it even more interesting and add a modifier that should go in the scene evoked by "judge".

Suppose we're talking about a dog show, and thus refer to a "dog judge" (somebody who judges the dogs in the competition).

-

[[the_E [tall_S (judge)_A]_E judge_C]_P dog_A (judge)_A]_A was able to reach... Is this too complex of a P?

[the_E [tall_S (judge)_A]_E [dog_A judge_P+A]_C]_A was able to reach... This feels like the more natural structure given the compositionality of the sentence. But it has a C-scene.

— You are receiving this because you were assigned. Reply to this email directly, view it on GitHub https://github.com/UniversalConceptualCognitiveAnnotation/docs/issues/25#issuecomment-427047963, or mute the thread https://github.com/notifications/unsubscribe-auth/AIG86xiNOr5PcOoeWXhXTjxrCCLNosebks5uhh_NgaJpZM4WoF_G .

nschneid commented 5 years ago

OK, so

But

?

omriabnd commented 5 years ago

Yes. Exactly.

On Thu, Oct 4, 2018 at 6:14 PM Nathan Schneider notifications@github.com wrote:

OK, so

  • the driver: [the_E driver_C]_P+A
  • the taxi driver: taxi_A [the_E driver_C]_P+A
  • the driver of the taxi: [the_E driver_C]_P+A [of_R the_E taxi_C]_A
  • the experienced taxi driver: experienced_D taxi_A [the_E driver_C]_P+A (presumably "experienced" relates to the manner of driving)

But

  • the tall taxi driver: the_E [tall_S (driver)_A]_E [taxi_A driver_P+A]_C
  • the taxi driver who swims: the_E [taxi_A driver_P+A]_C [who_R [swims_P (driver)_A]_C]_E

?

— You are receiving this because you were assigned. Reply to this email directly, view it on GitHub https://github.com/UniversalConceptualCognitiveAnnotation/docs/issues/25#issuecomment-427057486, or mute the thread https://github.com/notifications/unsubscribe-auth/AIG86zcUWTGO45GuV1g-C8jLb8PKIlTQks5uhiW8gaJpZM4WoF_G .

dotdv commented 5 years ago

added the option of "C-scenes" under subsection "Model 3: Inter-Scene relations" *mentioned that they should be used sparingly

nschneid commented 5 years ago

the experienced taxi driver: experienced_D taxi_A [the_E driver_C]_P+A (presumably "experienced" relates to the manner of driving)

A consequence of this is we don't really know whether the participant-entity or the scene itself is the head of the unit (implicitly, a normal Scene unit is headed by a P/S). Thus "the driver scared her" and "the driving scared her" would be structurally similar:

Are we OK with that? I guess the assumption could be that in a P+A unit, the A is the category that applies to the full word: "driver" as a whole refers to the person. Presumably zero-derived role words like "boss" and "engineer" would have P+A for the noun (entity) use and just P for the verb (event) use.

What about more complicated derivational morphology where a noun for a phenomenon or quality is derived from a participant: "consumerism", "criminality", "demagoguery", "musicianship"? Should these be simply P/S, ignoring the fact that they are derived because the participants are generic?

dotdv commented 5 years ago

added the option of "C-scenes" under subsection "Model 3: Inter-Scene relations" *mentioned that they should be used sparingly

So far we know that a role/profession should first be marked a C-scene if one of its siblings does not modify the process that the role evokes (tall judge or judge who swims).

But is it also true to say that if a role/profession is the head of a RC we necessarily have to mark the role C? Related streusle example: Davie_A is_S [[a_F [patient_C and_N methodical_C]_D teacher_P+A]_C [who has a great ear and sensitivity for his students passions]_E]_A?

And the next question would be what to do if the same scene appears as a top level scene? [patient_C and_N methodical_C]_D teacherP+A [who has a great ear...]\?]_H?

omriabnd commented 5 years ago

I agree with both analyses you suggested.

On Mon, Nov 19, 2018 at 9:58 AM dotdv notifications@github.com wrote:

added the option of "C-scenes" under subsection "Model 3: Inter-Scene relations" *mentioned that they should be used sparingly

So far we know that a role/profession should first be marked a C-scene if one of its siblings does not modify the process that the role evokes (tall judge or judge who swims).

But is it also true to say that if a role/profession is the head of a RC we necessarily have to mark the role C? Related streusle example: Davie_A is_S [[a_F [patient_C and_N methodical_C]_D teacher_P+A]_C [who has a great ear and sensitivity for his students passions]_E]_A?

And the next question would be what to do if the same scene appears as a top level scene? [patient_C and_N methodical_C]D teacher_P+A [who has a great ear...] ?]_H?

— You are receiving this because you were assigned. Reply to this email directly, view it on GitHub https://github.com/UniversalConceptualCognitiveAnnotation/docs/issues/25#issuecomment-439802411, or mute the thread https://github.com/notifications/unsubscribe-auth/AIG861yrbq-IbxHivgBSOOHYfIt3OX50ks5uwmSxgaJpZM4WoF_G .

nschneid commented 5 years ago

Davie_A is_S [[a_F [patient_C and_N methodical_C]_D teacher_P+A]_C [who has a great ear and sensitivity for his students passions]_E]_A?

Looks OK to me, though I worry we could get cases like "tall and patient teacher" where only one adjective describes the act of teaching.

I don't understand how the second example is different.