%0 Conference Paper %1 chen-bansal-2018-fast %A Chen, Yen-Chun %A Bansal, Mohit %B Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) %C Melbourne, Australia %D 2018 %E Gurevych, Iryna %E Miyao, Yusuke %I Association for Computational Linguistics %K pointer reinforcement-learning summarisation %P 675--686 %R 10.18653/v1/P18-1063 %T Fast Abstractive Summarization with Reinforce-Selected Sentence Rewriting %U https://aclanthology.org/P18-1063 %X Inspired by how humans summarize long documents, we propose an accurate and fast summarization model that first selects salient sentences and then rewrites them abstractively (i.e., compresses and paraphrases) to generate a concise overall summary. We use a novel sentence-level policy gradient method to bridge the non-differentiable computation between these two neural networks in a hierarchical way, while maintaining language fluency. Empirically, we achieve the new state-of-the-art on all metrics (including human evaluation) on the CNN/Daily Mail dataset, as well as significantly higher abstractiveness scores. Moreover, by first operating at the sentence-level and then the word-level, we enable parallel decoding of our neural generative model that results in substantially faster (10-20x) inference speed as well as 4x faster training convergence than previous long-paragraph encoder-decoder models. We also demonstrate the generalization of our model on the test-only DUC-2002 dataset, where we achieve higher scores than a state-of-the-art model.

@inproceedings{chen-bansal-2018-fast, abstract = {Inspired by how humans summarize long documents, we propose an accurate and fast summarization model that first selects salient sentences and then rewrites them abstractively (i.e., compresses and paraphrases) to generate a concise overall summary. We use a novel sentence-level policy gradient method to bridge the non-differentiable computation between these two neural networks in a hierarchical way, while maintaining language fluency. Empirically, we achieve the new state-of-the-art on all metrics (including human evaluation) on the CNN/Daily Mail dataset, as well as significantly higher abstractiveness scores. Moreover, by first operating at the sentence-level and then the word-level, we enable parallel decoding of our neural generative model that results in substantially faster (10-20x) inference speed as well as 4x faster training convergence than previous long-paragraph encoder-decoder models. We also demonstrate the generalization of our model on the test-only DUC-2002 dataset, where we achieve higher scores than a state-of-the-art model.}, added-at = {2024-04-16T13:13:18.000+0200}, address = {Melbourne, Australia}, author = {Chen, Yen-Chun and Bansal, Mohit}, biburl = {https://www.bibsonomy.org/bibtex/2b8333a2088a055760850007aad03ab35/antonvlasjuk}, booktitle = {Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)}, doi = {10.18653/v1/P18-1063}, editor = {Gurevych, Iryna and Miyao, Yusuke}, interhash = {969f40f48550abcfc3c0921a53254e1f}, intrahash = {b8333a2088a055760850007aad03ab35}, keywords = {pointer reinforcement-learning summarisation}, month = jul, pages = {675--686}, publisher = {Association for Computational Linguistics}, timestamp = {2024-04-16T13:13:18.000+0200}, title = {Fast Abstractive Summarization with Reinforce-Selected Sentence Rewriting}, url = {https://aclanthology.org/P18-1063}, year = 2018 }