{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/learning-to-extract-coherent-summary-via-deep","title":"Learning to Extract Coherent Summary via Deep Reinforcement Learning","arxiv_id":"1804.07036","date":"2018-04-19","proceeding":null,"authors":["Yuxiang Wu","Baotian Hu"],"abstract":"Coherence plays a critical role in producing a high-quality summary from a\ndocument. In recent years, neural extractive summarization is becoming\nincreasingly attractive. However, most of them ignore the coherence of\nsummaries when extracting sentences. As an effort towards extracting coherent\nsummaries, we propose a neural coherence model to capture the cross-sentence\nsemantic and syntactic coherence patterns. The proposed neural coherence model\nobviates the need for feature engineering and can be trained in an end-to-end\nfashion using unlabeled data. Empirical results show that the proposed neural\ncoherence model can efficiently capture the cross-sentence coherence patterns.\nUsing the combined output of the neural coherence model and ROUGE package as\nthe reward, we design a reinforcement learning method to train a proposed\nneural extractive summarizer which is named Reinforced Neural Extractive\nSummarization (RNES) model. The RNES model learns to optimize coherence and\ninformative importance of the summary simultaneously. Experimental results show\nthat the proposed RNES outperforms existing baselines and achieves\nstate-of-the-art performance in term of ROUGE on CNN/Daily Mail dataset. The\nqualitative evaluation indicates that summaries produced by RNES are more\ncoherent and readable.","url_abs":"http://arxiv.org/abs/1804.07036v1","url_pdf":"http://arxiv.org/pdf/1804.07036v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[],"tasks":[{"task_slug":"deep-reinforcement-learning","task_name":"Deep Reinforcement Learning"},{"task_slug":"extractive-summarization","task_name":"Extractive Summarization"},{"task_slug":"feature-engineering","task_name":"Feature Engineering"},{"task_slug":"reinforcement-learning","task_name":"Reinforcement Learning"},{"task_slug":"reinforcement-learning-1","task_name":"Reinforcement Learning (RL)"},{"task_slug":"sentence","task_name":"Sentence"},{"task_slug":"text-summarization","task_name":"Text Summarization"},{"task_slug":"reinforcement-learning-2","task_name":"reinforcement-learning"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/text-summarization-on-cnn-daily-mail-2","task":"Text Summarization","dataset":"CNN / Daily Mail (Anonymized)","model":"RNES w/o coherence","rank_in_archive_order":3,"of":13,"metrics":{"ROUGE-1":"41.25","ROUGE-2":"18.87","ROUGE-L":"37.75"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1804.07036","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}