AMR Parsing is Far from Solved: GrAPES, the Granular AMR Parsing Evaluation Suite

We present the Granular AMR Parsing Evaluation Suite (GrAPES), a challenge set for Abstract Meaning Representation (AMR) parsing with accompanying evaluation metrics. AMR parsers now obtain high scores on the standard AMR evaluation metric Smatch, close to or even above reported inter-annotator agreement. But that does not mean that AMR parsing is solved; in fact, human evaluation in previous work indicates that current parsers still quite frequently make errors on node labels or graph structure that substantially distort sentence meaning. Here, we provide an evaluation suite that tests AMR parsers on a range of phenomena of practical, technical, and linguistic interest. Our 36 categories range from seen and unseen labels, to structural generalization, to coreference. GrAPES reveals in depth the abilities and shortcomings of current AMR parsers. Set Category Metric AM Parser C&L AMRBart # 1 Pragmatic coreference (testset) Edge recall 06 [02, 18] 08 [03, 22] 39 [25, 55] 36 Prerequisites 50 [34, 66] 36 [22, 52] 61 [45, 75] 36 Pragmatic coreference (Winograd) Edge recall 02 [00, 13] 05 [01, 17] 32 [20, 48] 40 Prerequisites 78 [62, 88] 30 [18, 45] 65 [50, 78] 40 2 Syntactic (gap) reentrancies Edge recall 24 [14, 39] 24 [14, 39] 49 [34, 64] 41 Prerequisites 54 [39, 68] 59 [43, 72] 68 [53, 80] 41 Unambiguous coreference Edge recall 10 [03, 25] 39 [24, 56] 65 [47, 79] 31 Prerequisites 71 [53, 84] 71 [53, 84] 77 [60, 89] 31