amrisi / amr-guidelines

250 stars 87 forks source link

"in light of...her alleged role in the party's graft scandal" #132

Open nschneid opened 9 years ago

nschneid commented 9 years ago

A tricky sentence from Haaretz:

Yisrael Beiteinu Knesset member and Deputy Interior Minister Faina Kirshenbaum asked party chairman Avigdor Lieberman not to seek re-election in light of the police investigation into her alleged role in the party's graft scandal.

(This actually means that Kirshenbaum asked Lieberman for permission not to run for reelection. The gender of the pronoun is important to disambiguating who is under investigation.)

My attempt:

(a / ask-02
      :ARG0 (w / woman :name (n / name :op1 "Faina" :op2 "Kirshenbaum")
            :ARG0-of (h / have-org-role-91
                  :ARG1 (p / parliament :name (n2 / name :op1 "Knesset"))
                  :ARG2 (m / member))
            :ARG0-of (h2 / have-org-role-91
                  :ARG1 (p2 / political-party :name (n3 / name :op1 "Yisrael" :op2 "Beiteinu")))
            :ARG0-of (h3 / have-org-role-91
                  :ARG2 (m2 / minister)
                  :ARG3 (i / interior)
                  :mod (d / deputy)))
      :ARG1 (s / seek-01 :polarity -
            :ARG0 w
            :ARG1 (r / reelect-01
                  :ARG1 w))
      :ARG2 (p3 / person :name (n4 / name :op1 "Avigdor" :op2 "Lieberman")
            :ARG0-of (h4 / have-org-role-91
                  :ARG1 p2
                  :ARG2 (c / chairman)))
      :given (i2 / investigate-01
            :ARG0 (p4 / police)
            :ARG1 w
            :ARG2 (r2 / role
                  :topic (s3 / scandal-u-01
                        :ARG0 p2
                        :ARG1 (g / graft-00
                              :ARG0 p2))
                  :ARG1-of (a2 / allege-01))))

The most difficult part was at the end:

uhermjakob commented 9 years ago

An interesting sentence indeed, Nathan.

Yisrael Beiteinu Knesset member and Deputy Interior Minister Faina Kirshenbaum asked party chairman Avigdor Lieberman not to seek re-election in light of the police investigation into her alleged role in the party's graft scandal.

(a / ask-02
      :ARG0 (w / woman :name (n / name :op1 "Faina" :op2 "Kirshenbaum")
            :ARG0-of (h / have-org-role-91
                  :ARG1 (p / parliament :name (n2 / name :op1 "Knesset"))
                  :ARG2 (m / member))
            :ARG0-of (h2 / have-org-role-91
                  :ARG1 (p2 / political-party :name (n3 / name :op1 "Yisrael" :op2 "Beiteinu")))
            :ARG0-of (h3 / have-org-role-91
                  :ARG2 (m2 / minister
                        :topic (i / interior)
                        :mod (d / deputy))))
      :ARG1 (s / seek-01 :polarity -
            :ARG0 w
            :ARG1 (r / reelect-01
                  :ARG1 w))
      :ARG2 (p3 / person :name (n4 / name :op1 "Avigdor" :op2 "Lieberman")
            :ARG0-of (h4 / have-org-role-91
                  :ARG1 p2
                  :ARG2 (c / chairman)))
      :ARG1-of (c2 / cause-01
            :ARG0 (i2 / investigate-01
                  :ARG0 (p4 / police)
                  :ARG1 w
                  :ARG2 (a3 / act-01
                        :ARG0 w
                        :ARG1 (s2 / scandalize-01
                              :ARG1 p2
                              :ARG2 (g / graft))
                        :ARG1-of (a2 / allege-01)))))
cbonial commented 9 years ago

Regarding the use of "act-01" (as Ulf suggests) above vs. creating a "role" roleset (as Nathan suggests): I think this brings up an important issue where the AMR group needs to decide whether we generally want to... 1) rely on the creation of new rolesets or 2) rely on suggestions from the tool, pointing to an existing roleset ...when new events/states are encountered that don't have a clear, etymologically related counterpart with an existing roleset. To some extent, I think the addition of new rolesets in these cases would be a solution that's more transparent to outside users of the AMR corpus, and it avoids the potential loss of semantic specificity that may come along with mapping to etymologically unrelated predicates. Additionally, encouraging annotators to consider whether or not a word needs a roleset (use role-00), as opposed to encouraging annotators to consider a new mapping to a semantically similar predicate, may lead to higher IAA. On the other hand, this type of mapping (as opposed to roleset creation) could be seen as a type of semantic decomposition of "role," which seems to be in line with the goals of AMR.
I simply think that we should consider what general tendency we want to follow in these cases, as this issue will surely continue to arise: new rolesets, or new tool suggestions mapping to existing rolesets.