{"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/unsupervised-text-style-transfer-using","title":"Unsupervised Text Style Transfer using Language Models as Discriminators","arxiv_id":"1805.11749","date":"2018-05-30","proceeding":"NeurIPS 2018 12","authors":["Zichao Yang","Zhiting Hu","Chris Dyer","Eric P. Xing","Taylor Berg-Kirkpatrick"],"abstract":"Binary classifiers are often employed as discriminators in GAN-based\nunsupervised style transfer systems to ensure that transferred sentences are\nsimilar to sentences in the target domain. One difficulty with this approach is\nthat the error signal provided by the discriminator can be unstable and is\nsometimes insufficient to train the generator to produce fluent language. In\nthis paper, we propose a new technique that uses a target domain language model\nas the discriminator, providing richer and more stable token-level feedback\nduring the learning process. We train the generator to minimize the negative\nlog likelihood (NLL) of generated sentences, evaluated by the language model.\nBy using a continuous approximation of discrete sampling under the generator,\nour model can be trained using back-propagation in an end- to-end fashion.\nMoreover, our empirical results show that when using a language model as a\nstructured discriminator, it is possible to forgo adversarial steps during\ntraining, making the process more stable. We compare our model with previous\nwork using convolutional neural networks (CNNs) as discriminators and show that\nour approach leads to improved performance on three tasks: word substitution\ndecipherment, sentiment modification, and related language translation.","url_abs":"http://arxiv.org/abs/1805.11749v3","url_pdf":"http://arxiv.org/pdf/1805.11749v3.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":"unsupervised-text-style-transfer-using","repo_url":"https://github.com/asyml/texar/tree/master/examples/text_style_transfer","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null}],"tasks":[{"task_slug":"decipherment","task_name":"Decipherment"},{"task_slug":"language-modeling","task_name":"Language Modeling"},{"task_slug":"language-modelling","task_name":"Language Modelling"},{"task_slug":"style-transfer","task_name":"Style Transfer"},{"task_slug":"text-style-transfoer","task_name":"Text Style Transfer"},{"task_slug":"translation","task_name":"Translation"},{"task_slug":"unsupervised-text-style-transfer","task_name":"Unsupervised Text Style Transfer"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1805.11749","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}