Probabilistic, Structure-Aware Algorithms for Improved Variety, Accuracy, and Coverage of AMR Alignments

TL;DR

This paper introduces probabilistic, structure-aware algorithms for AMR alignment that improve coverage, accuracy, and variety by combining unsupervised learning with heuristics, without needing syntactic parses.

Contribution

It presents novel algorithms that leverage unsupervised models sensitive to graph substructures, enhancing AMR alignment coverage and accuracy over previous methods.

Findings

Higher coverage of AMR nodes and edges

Improved alignment accuracy

Broader variety of AMR substructures covered

Abstract

We present algorithms for aligning components of Abstract Meaning Representation (AMR) graphs to spans in English sentences. We leverage unsupervised learning in combination with heuristics, taking the best of both worlds from previous AMR aligners. Our unsupervised models, however, are more sensitive to graph substructures, without requiring a separate syntactic parse. Our approach covers a wider variety of AMR substructures than previously considered, achieves higher coverage of nodes and edges, and does so with higher accuracy. We will release our LEAMR datasets and aligner for use in research on AMR parsing, generation, and evaluation.