AMR comparation


I am trying to find references comparing MRS with AMR. I found some papers about AMR expressivity like but none comparing MRS, RMRS or DMRS with AMR. Suggestions?


There’s a brief comparison in my dissertation around p38:

Bender et al. 2015 ( compare more deeply, looking at the benefits of compositionality.

Also see, and I think Stephan’s group may have done some other comparisons but I’m not sure which publications they’re in.


Buys and Blunsom (2017) also briefly compare them, concluding that DMRS resources are more consistent than AMR resources:


Yes, the paper Bender et al. 2015 is quite nice. But I didn’t quite follow all the discussions of section 2, deep issues very condensed. I missed more examples. Hi @ebender, if you have any other material with a more didactic presentation about compositionality I would be happy to read! :wink:

we explore which aspects of meaning among those captured by annotation projects serving NLP work (and thus presumably of interest to the NLP community) can be seen as compositional.


It’s possible that something from the book I’m currently working on with Alex L. might help, but can you articulate any specific questions that section 2 leaves you with?


Oh, I also usually have a hard time reading about some of the semantic issues, I think often just because of the proportion of terminology…

Note, however,
that there may be disagreement as to whether particular linkings (e.g. between the subject of a participial
modifier and the subject of the clause it modifies) are required by the grammar or simply anaphoric in

What would be an example of the disagreement here? (That would be one place where I would benefit from an example, perhaps that does not match Alexandre’s).


The parenthetical there is an example of a disagreement:

(e.g. between the subject of a participial
modifier and the subject of the clause it modifies)

Here’s a linguistic example of that phenomenon:

(1) Having eaten, Kim left the table.
(2) Pickled to just the right amount, Kim is sure these cucumbers will taste great.

Many theories have necessary reentrancy between the missing subject of the participial modifier (here Having eaten and Pickled to just the right amount). But Dan refuses this analysis because of examples like (2) where the sensible reading of the sentence doesn’t have them coreferential.


Btw, for some cross-framework discussion of this particular example, see: