Global Encoding for Abstractive Summarization

TL;DR

This paper introduces a global encoding framework for neural abstractive summarization that enhances source representations and reduces repetition, leading to improved performance on benchmark datasets.

Contribution

The paper presents a novel global encoding method with a convolutional gated unit to control information flow, addressing repetition and semantic irrelevance in seq2seq models.

Findings

Outperforms baseline models on LCSTS and Gigaword datasets

Reduces repetition in generated summaries

Improves semantic relevance of summaries

Abstract

In neural abstractive summarization, the conventional sequence-to-sequence (seq2seq) model often suffers from repetition and semantic irrelevance. To tackle the problem, we propose a global encoding framework, which controls the information flow from the encoder to the decoder based on the global information of the source context. It consists of a convolutional gated unit to perform global encoding to improve the representations of the source-side information. Evaluations on the LCSTS and the English Gigaword both demonstrate that our model outperforms the baseline models, and the analysis shows that our model is capable of reducing repetition.