Text Summarization With Graph Attention Networks

TL;DR

This paper explores the use of graph attention networks and RST/Co-reference graphs to improve text summarization, but finds that simpler models perform better, and introduces a new annotated dataset for future research.

Contribution

It demonstrates that graph attention networks may not outperform simpler models in summarization and provides a new annotated dataset with RST graphs for benchmarking.

Findings

Graph attention networks did not improve summarization performance.

A simple MLP architecture outperformed the graph-based models.

Annotated XSum with RST graphs to establish a new benchmark.

Abstract

This study aimed to leverage graph information, particularly Rhetorical Structure Theory (RST) and Co-reference (Coref) graphs, to enhance the performance of our baseline summarization models. Specifically, we experimented with a Graph Attention Network architecture to incorporate graph information. However, this architecture did not enhance the performance. Subsequently, we used a simple Multi-layer Perceptron architecture, which improved the results in our proposed model on our primary dataset, CNN/DM. Additionally, we annotated XSum dataset with RST graph information, establishing a benchmark for future graph-based summarization models. This secondary dataset posed multiple challenges, revealing both the merits and limitations of our models.