Investigating the Effect of Relative Positional Embeddings on AMR-to-Text Generation with Structural Adapters
This work addresses graph-to-text generation for NLP researchers, but it is incremental as it builds on existing methods like StructAdapt.
The paper investigates the effect of Relative Positional Embeddings (RPE) on AMR-to-text generation and tests the robustness of StructAdapt, revealing that RPE may partially encode input graphs.
Text generation from Abstract Meaning Representation (AMR) has substantially benefited from the popularized Pretrained Language Models (PLMs). Myriad approaches have linearized the input graph as a sequence of tokens to fit the PLM tokenization requirements. Nevertheless, this transformation jeopardizes the structural integrity of the graph and is therefore detrimental to its resulting representation. To overcome this issue, Ribeiro et al. have recently proposed StructAdapt, a structure-aware adapter which injects the input graph connectivity within PLMs using Graph Neural Networks (GNNs). In this paper, we investigate the influence of Relative Position Embeddings (RPE) on AMR-to-Text, and, in parallel, we examine the robustness of StructAdapt. Through ablation studies, graph attack and link prediction, we reveal that RPE might be partially encoding input graphs. We suggest further research regarding the role of RPE will provide valuable insights for Graph-to-Text generation.