{"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/abstractive-text-summarization-by","title":"Abstractive Text Summarization by Incorporating Reader Comments","arxiv_id":"1812.05407","date":"2018-12-13","proceeding":null,"authors":["Shen Gao","Xiuying Chen","Piji Li","Zhaochun Ren","Lidong Bing","Dongyan Zhao","Rui Yan"],"abstract":"In neural abstractive summarization field, conventional sequence-to-sequence\nbased models often suffer from summarizing the wrong aspect of the document\nwith respect to the main aspect. To tackle this problem, we propose the task of\nreader-aware abstractive summary generation, which utilizes the reader comments\nto help the model produce better summary about the main aspect. Unlike\ntraditional abstractive summarization task, reader-aware summarization\nconfronts two main challenges: (1) Comments are informal and noisy; (2) jointly\nmodeling the news document and the reader comments is challenging. To tackle\nthe above challenges, we design an adversarial learning model named\nreader-aware summary generator (RASG), which consists of four components: (1) a\nsequence-to-sequence based summary generator; (2) a reader attention module\ncapturing the reader focused aspects; (3) a supervisor modeling the semantic\ngap between the generated summary and reader focused aspects; (4) a goal\ntracker producing the goal for each generation step. The supervisor and the\ngoal tacker are used to guide the training of our framework in an adversarial\nmanner. Extensive experiments are conducted on our large-scale real-world text\nsummarization dataset, and the results show that RASG achieves the\nstate-of-the-art performance in terms of both automatic metrics and human\nevaluations. The experimental results also demonstrate the effectiveness of\neach module in our framework. We release our large-scale dataset for further\nresearch.","url_abs":"http://arxiv.org/abs/1812.05407v1","url_pdf":"http://arxiv.org/pdf/1812.05407v1.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":"abstractive-text-summarization","task_name":"Abstractive Text Summarization"},{"task_slug":"reader-aware-summarization","task_name":"Reader-Aware Summarization"},{"task_slug":"text-summarization","task_name":"Text Summarization"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/reader-aware-summarization-on-rasg","task":"Reader-Aware Summarization","dataset":"RASG","model":"RASG","rank_in_archive_order":1,"of":1,"metrics":{"ROUGE-1":"30.33"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1812.05407","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}