Structural Information Preserving for Graph-to-Text Generation

TL;DR

This paper introduces a multi-view autoencoding approach with additional training signals to improve the preservation of structural information in graph-to-text generation, outperforming existing models on benchmark datasets.

Contribution

It proposes a novel multi-view autoencoding training method that enhances structural preservation in graph-to-text generation models.

Findings

Significant improvement over baseline models on benchmark datasets.

Effective preservation of input graph structure in generated text.

Multi-view autoencoding enhances model understanding of graph structures.

Abstract

The task of graph-to-text generation aims at producing sentences that preserve the meaning of input graphs. As a crucial defect, the current state-of-the-art models may mess up or even drop the core structural information of input graphs when generating outputs. We propose to tackle this problem by leveraging richer training signals that can guide our model for preserving input information. In particular, we introduce two types of autoencoding losses, each individually focusing on different aspects (a.k.a. views) of input graphs. The losses are then back-propagated to better calibrate our model via multi-task training. Experiments on two benchmarks for graph-to-text generation show the effectiveness of our approach over a state-of-the-art baseline. Our code is available at \url{http://github.com/Soistesimmer/AMR-multiview}.