âDonât Give Me the Details, Just the Summary! Topic-Aware Convolutional Neural Networks for Extreme Summarizationâ, 2018-08-27 (; backlinks; similar)â :
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 âWhat is the article about?â.
We collect a real-world, large-scale dataset XSum for this task by harvesting online articles from the British Broadcasting Corporation (BBC). We propose a novel abstractive model which is conditioned on the articleâs 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.