The content of an image/thing displayed

[mgeorgescu]

Any preferred choice for the following?

image of the 3 other criminals (i / image :mod (c2 / criminal (i / image :topic (c2 / criminal

the image with 4 criminals (i / image :prep-with (c2 / criminal (i / image :mod (c2 / criminal i / image :topic (c2 / criminal (i / image :ARG0-of (d2 / display-01 :ARG1 (c / criminal)))

We would opt for :topic. Thank you.

[uhermjakob]

I support :topic.

In NomBank, which we consider as a source for possible future frames for nouns, that role is :ARG1 ("theme"), the same role that Nombank uses for the topic of a book, for which many annotators have already been using :topic in AMR.

[nschneid]

The possibilities I would consider:

• Start using NomBank frames.
• Develop a generic relation frame between entities for cases where the non-core roles aren't specific enough. This was briefly discussed in #31. Something like:

(h / has-relation-91
  :ARG1 (t / thing)
  :ARG2 (c / criminal)
  :ARG3 (i / image))

This is spiritually akin to saying (t / thing :image-of (c / criminal)), but would keep the official non-core role set to a manageable number. (If we want to introduce NomBank frames at a later time, this solution would make it clear that such a frame would be desirable here.)

• Use :topic, though it's not a great fit. (Or rename the role as :content.)
• Decide on a verb frame: for "image" or "picture" I would favor depict-01 over display-01.