Don't Give Me the Details, Just the Summary! Topic-Aware Convolutional Neural Networks for Extreme Summarization
This addresses the problem of generating concise, abstractive summaries for news articles, which is incremental as it builds on existing summarization methods with a novel model.
The authors tackled extreme summarization, a new task requiring abstractive one-sentence summaries of news articles, by proposing a topic-aware convolutional neural network model that outperformed oracle extractive and state-of-the-art abstractive systems in automatic and human evaluations.
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.