Online Back-Parsing for AMR-to-Text Generation

TL;DR

This paper introduces an online back-parsing decoder for AMR-to-text generation that improves meaning preservation by jointly predicting AMR graphs during text generation, outperforming previous models.

Contribution

It presents a novel decoder that back predicts AMR graphs during generation, enhancing semantic fidelity over standard language modeling decoders.

Findings

Outperforms previous state-of-the-art on two AMR benchmarks

Better preservation of input meaning in generated text

Demonstrates effectiveness of back-parsing approach

Abstract

AMR-to-text generation aims to recover a text containing the same meaning as an input AMR graph. Current research develops increasingly powerful graph encoders to better represent AMR graphs, with decoders based on standard language modeling being used to generate outputs. We propose a decoder that back predicts projected AMR graphs on the target sentence during text generation. As the result, our outputs can better preserve the input meaning than standard decoders. Experiments on two AMR benchmarks show the superiority of our model over the previous state-of-the-art system based on graph Transformer.