Giters

amrisi / amr-guidelines

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

on the meaning of quotation marks

flipz357 opened this issue · comments

commented

I am sorry if this has been discussed before, but I could not find something.

Background: I notice some little inconsistencies w.r.t. quotation mark usage. For instance, in LDC17, once "Route 288, the circumferential highway...." is projected onto

:ARG1 (h / highway :wiki "Virginia_State_Route_288"
            :name (r / name :op1 "Route" :op2 288)

i.e., op2 288.

And in a slightly different context ("288, for those of us that actually use it..."):

:ARG1 (r / road :wiki "Virginia_State_Route_288" :name (x / name :op1 "288"))

i.e., op1 "288".

Here are my questions/thoughts:

  • I think it should be with quotation marks in both cases. Is this correct?
  • It has been speculated that usage of quotation marks in AMR is just "sugar/redundant" and that it is just to be more compliant with Penman language. Is this true? Maybe the guidelines could comment on this (or I missed something...)? I personally like the small, but, in my opinion, useful meaning distinction between stringified constants and constants that, e.g., represent pure numbers etc.
  • In the referenced issue it is also speculated that it should be easy to create a deterministic map from any AMR without quotes to an AMR that contains stringified nodes (a lossless reconstruction would really imply that the quotation marks are just... sugar). But I am not so sure if it is really possible.
commented

Numerals consisting digits and optionally a decimal point are recognized by the AMR Editor, which is convenient for entering quantities and values. But note that negative numbers, for example, need to be quoted. Consider the dictionary example:

(have-degree-91 :li 7 :li "-1" :ARG1 algorithm :ARG2 fast :ARG3 more)

So I would say in a normalized form they should all be quoted, i.e.

(have-degree-91 :li "7" :li "-1" :ARG1 algorithm :ARG2 fast :ARG3 more)

but in any case I don't know that Penman readers treat numeric constants any differently from string constants.

commented

Thanks a lot! I think quoting negative numbers also makes sense as a means of escaping the special "-" sign (e.g., in run-01, where - indicates "now comes the sense").

but in any case I don't know that Penman readers treat numeric constants any differently from string constants.

Yes, I think it treats them the same. I just wondered, whether in AMR, the quotes carry some "meaning", in the sense that "" is like a predicate P(value) := "value" and I just don't know what P really means. But I suppose that's not really the case, or is it?

commented

Closing this since I don't know myself anymore what that issue was about