{"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/adversarial-ranking-for-language-generation","title":"Adversarial Ranking for Language Generation","arxiv_id":"1705.11001","date":"2017-05-31","proceeding":"NeurIPS 2017 12","authors":["Kevin Lin","Dianqi Li","Xiaodong He","Zhengyou Zhang","Ming-Ting Sun"],"abstract":"Generative adversarial networks (GANs) have great successes on synthesizing\ndata. However, the existing GANs restrict the discriminator to be a binary\nclassifier, and thus limit their learning capacity for tasks that need to\nsynthesize output with rich structures such as natural language descriptions.\nIn this paper, we propose a novel generative adversarial network, RankGAN, for\ngenerating high-quality language descriptions. Rather than training the\ndiscriminator to learn and assign absolute binary predicate for individual data\nsample, the proposed RankGAN is able to analyze and rank a collection of\nhuman-written and machine-written sentences by giving a reference group. By\nviewing a set of data samples collectively and evaluating their quality through\nrelative ranking scores, the discriminator is able to make better assessment\nwhich in turn helps to learn a better generator. The proposed RankGAN is\noptimized through the policy gradient technique. Experimental results on\nmultiple public datasets clearly demonstrate the effectiveness of the proposed\napproach.","url_abs":"http://arxiv.org/abs/1705.11001v3","url_pdf":"http://arxiv.org/pdf/1705.11001v3.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":[{"paper_slug":"adversarial-ranking-for-language-generation","repo_url":"https://github.com/desire2020/RankGAN","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"tf","reach":null}],"tasks":[{"task_slug":null,"task_name":"Generative Adversarial Network"},{"task_slug":"text-generation","task_name":"Text Generation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/text-generation-on-coco-captions","task":"Text Generation","dataset":"COCO Captions","model":"RankGAN","rank_in_archive_order":3,"of":5,"metrics":{"BLEU-2":"0.850","BLEU-3":"0.672","BLEU-4":"0.557","BLEU-5":"0.544"},"uses_additional_data":false},{"leaderboard":"/sota/text-generation-on-chinese-poems","task":"Text Generation","dataset":"Chinese Poems","model":"RankGAN","rank_in_archive_order":1,"of":3,"metrics":{"BLEU-2":"0.812"},"uses_additional_data":false},{"leaderboard":"/sota/text-generation-on-emnlp2017-wmt","task":"Text Generation","dataset":"EMNLP2017 WMT","model":"RankGAN","rank_in_archive_order":5,"of":5,"metrics":{"BLEU-2":"0.778","BLEU-3":"0.478","BLEU-4":"0.411","BLEU-5":"0.463"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1705.11001","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}