Maximum Bayes Smatch Ensemble Distillation for AMR Parsing

TL;DR

This paper introduces a novel ensemble distillation method for AMR parsing that combines Smatch-based ensembling to significantly improve performance across multiple languages and domains, surpassing previous state-of-the-art results.

Contribution

It proposes a maximum Bayes Smatch ensemble distillation technique that effectively leverages ensemble methods to enhance AMR parser accuracy, especially in high-performing models.

Findings

Achieved new state-of-the-art scores for English AMR parsing (85.9 and 84.3 for AMR2.0 and AMR3.0).

Restored substantial gains from silver data augmentation in high-performance models.

Attained new state-of-the-art results for cross-lingual AMR parsing and domain adaptation.

Abstract

AMR parsing has experienced an unprecendented increase in performance in the last three years, due to a mixture of effects including architecture improvements and transfer learning. Self-learning techniques have also played a role in pushing performance forward. However, for most recent high performant parsers, the effect of self-learning and silver data augmentation seems to be fading. In this paper we propose to overcome this diminishing returns of silver data by combining Smatch-based ensembling techniques with ensemble distillation. In an extensive experimental setup, we push single model English parser performance to a new state-of-the-art, 85.9 (AMR2.0) and 84.3 (AMR3.0), and return to substantial gains from silver data augmentation. We also attain a new state-of-the-art for cross-lingual AMR parsing for Chinese, German, Italian and Spanish. Finally we explore the impact of the proposed technique on domain adaptation, and show that it can produce gains rivaling those of human annotated data for QALD-9 and achieve a new state-of-the-art for BioAMR.