Structural Adapters in Pretrained Language Models for AMR-to-text Generation

TL;DR

This paper introduces StructAdapt, a novel adapter method that efficiently encodes graph structure into pretrained language models for AMR-to-text generation, improving performance while training minimal additional parameters.

Contribution

StructAdapt effectively models graph node interactions within PLMs by training only graph structure-aware adapters, preserving language knowledge and enhancing graph-to-text generation.

Findings

Outperforms state-of-the-art on two AMR-to-text datasets

Requires training only 5.1% of PLM parameters

Explicit graph structure encoding improves generation quality

Abstract

Pretrained language models (PLM) have recently advanced graph-to-text generation, where the input graph is linearized into a sequence and fed into the PLM to obtain its representation. However, efficiently encoding the graph structure in PLMs is challenging because such models were pretrained on natural language, and modeling structured data may lead to catastrophic forgetting of distributional knowledge. In this paper, we propose StructAdapt, an adapter method to encode graph structure into PLMs. Contrary to prior work, StructAdapt effectively models interactions among the nodes based on the graph connectivity, only training graph structure-aware adapter parameters. In this way, we incorporate task-specific knowledge while maintaining the topological structure of the graph. We empirically show the benefits of explicitly encoding graph structure into PLMs using StructAdapt, outperforming the state of the art on two AMR-to-text datasets, training only 5.1% of the PLM parameters.