Pushing the Limits of AMR Parsing with Self-Learning

TL;DR

This paper demonstrates that self-learning techniques, including synthetic data generation and action refinement, can significantly improve AMR parsing performance, achieving state-of-the-art results without extra human annotations.

Contribution

It introduces self-learning methods for AMR parsing that enhance existing models and set new performance benchmarks without additional human annotation effort.

Findings

Improved AMR parser performance with self-learning techniques.

Achieved state-of-the-art results on AMR 1.0 and 2.0 datasets.

Self-learning methods outperform previous approaches without extra annotations.

Abstract

Abstract Meaning Representation (AMR) parsing has experienced a notable growth in performance in the last two years, due both to the impact of transfer learning and the development of novel architectures specific to AMR. At the same time, self-learning techniques have helped push the performance boundaries of other natural language processing applications, such as machine translation or question answering. In this paper, we explore different ways in which trained models can be applied to improve AMR parsing performance, including generation of synthetic text and AMR annotations as well as refinement of actions oracle. We show that, without any additional human annotations, these techniques improve an already performant parser and achieve state-of-the-art results on AMR 1.0 and AMR 2.0.