Don't Give Me the Details, Just the Summary! Topic-Aware Convolutional Neural Networks for Extreme Summarization

TL;DR

This paper introduces extreme summarization, a new abstractive summarization task for creating concise one-sentence news summaries, and proposes a CNN-based model conditioned on article topics that outperforms existing methods.

Contribution

The paper presents a novel extreme summarization task, a large-scale dataset, and a topic-aware convolutional neural network model that effectively captures long-range dependencies.

Findings

The CNN-based model outperforms oracle extractive systems.

The model surpasses state-of-the-art abstractive approaches.

Human evaluations favor the proposed method.

Abstract

We introduce extreme summarization, a new single-document summarization task which does not favor extractive strategies and calls for an abstractive modeling approach. The idea is to create a short, one-sentence news summary answering the question "What is the article about?". We collect a real-world, large-scale dataset for this task by harvesting online articles from the British Broadcasting Corporation (BBC). We propose a novel abstractive model which is conditioned on the article's topics and based entirely on convolutional neural networks. We demonstrate experimentally that this architecture captures long-range dependencies in a document and recognizes pertinent content, outperforming an oracle extractive system and state-of-the-art abstractive approaches when evaluated automatically and by humans.