:ARG0 for build-01 and move-01

[saramosher888]

The expanded AMR checker is also signaling an error if there is a country in :ARG0 for build-01 or move-01. For most of the concepts we are using, ARG0 is the agent/country. From the annotator's perspective, it seems more confusing than necessary to make ARG0 different for these two concepts than it is for all the others. In the argument structures listed in the main AMR dictionary, build-01 has "builder" as :ARG0, and move-01 has "mover" as :ARG0. So is there a reason why are are not supposed to use slots that already exist for this purpose?

I should mention that the frame key that I have been using as my regular reference (which is based on the frame pop-up that shows in AMR Editor) has :ARG0 as the agent / mover / builder. It seems that we did not catch the discrepancy between that and the Diplomacy-specific dictionary that Ulf is maintaining. So up to this point, all of the 1,000+ sentences that have been annotated have the country in :ARG0 for build-01 and move-01. In other words, the annotation is consistent--but it's consistent about filling in :ARG0 instead of leaving it blank. It is certainly possible to go back and change them all, but that's expensive in terms of time if they all need to be changed one by one.

So is it important to leave :ARG0 blank? If it is important, is it possible to write some code that can change all of them automatically?

[saramosher888]

One more comment: If we do decide to fill :ARG0, we will have a problem if we try to require :ARG0 to be filled. Likewise, we will have a problem if we try to require the unit to be modified with the country name, because sometimes that information is not available. Most of the time we know who the country is, but not always. In these examples, we know the original location of the unit that's moving, but we don't know what country owns the unit. This is not the majority of sentences, but it's fairly common.

Example:

• "I shoulda gone with my gut last turn to support ANK-CON."
• "Would you be willing to support either Gal or Rum into Bud from Vie?"

[uhermjakob]

Hi Sara,

• The expanded AMR checker should not and does not signal an error for any build-01 :ARG0 country. I just double-checked. build-01 :ARG0 country is typical, as included in the guidelines for build-01.
• For user smosher, there are 432 instances of move-01, of which 316 have an :ARG0. That is a majority but not anywhere close to all cases. That suggests that the powers for move-01 are actually not annotated consistently. Example: first move-01 in dip_train_0001.24
• I strongly recommend we stick with the current guidelines and do not annotate the power in two places for move-01 and similar frames.
• We will of course not require an :ARG0 for move-01, retreat-01 etc., which would be contrary to our own guidelines. We will also not require a :mod under unit for the reasons you mention.
• I realize that the corrections will not be fun, but my estimate for correcting the :ARG0s of move-01 is that it should take about 2 hours. (If you find it to be much slower, we can have a Zoom meeting to make sure you use the AMR Checker and AMR Editor most efficiently.) It is theoretically possible to write code to do that, but automatically fiddling with human-annotated AMRs has to be done with extreme care and much testing, which would exceed manual corrections in effort, I am afraid.

[saramosher888]

Hi Ulf, Thanks for your response. I agree that it will not be hard to remove :arg0, if that is truly what we want to do. Before I do that, I just want to understand the reasoning, especially as we're looking forward to hiring and training new annotators in the coming weeks. I am thinking about how to streamline processes in order to get people up and running faster and minimize potential mistakes. One way to do this is to make the right choice also the easiest / most straightforward thing. My question is this: if the policy is to add a "unit" in order to make a place to input the power, then why would we not put the power in the slot that already exists (:arg0)? I'm not intending this as criticism, but I would like to understand the reason. Is it because it's more important in some way to describe the unit than it is to specify the agent causing the move? Are we more interested in one piece of information than we are in the other? (I agree that it's a pain to duplicate it in both places.)

[uhermjakob]

Hi Sara, Entities are typically very important. For Diplomacy, these includes powers, provinces, seas and military units. Military units are easily identified by their type (army/fleet), possessor/nationality and location. Possession/nationality is an important and persistent characteristic of a unit, independent of any actions that might take place occasionally. So we might refer to a unit as the "French army in Paris" (army :mod France :location Paris). It might be involved in situations without a clear agent, e.g. in a description of the current state, or some bouncing. This "unit perspective" is reflected in having (optional) slots :mod and :location under the unit. That also mirrors the entityhood of a unit in DAIDE, where the nationality and location are sub-components of the unit. So we have been using the (army :mod France :location Paris) structure from the very beginning of Diplomacy. We later had some discussion regarding build-01, which is a bit different, because the unit to be built initially does not exist yet, it is only a result, so we use (build-01 :ARG0 country :ARG1 unit :location province/sea) instead, after some spirited discussion. With the nationality of the unit already described under the :mod of the unit, adding it also as the :ARG0 of move-01 is redundant as it does not really add a new semantic role.

[saramosher888]

Thanks, Ulf. It helps to know the thinking behind the decision. I'll work on making those changes to ARG0, and I'll talk to Jillian to make sure she is doing the same.