{"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/men-also-like-shopping-reducing-gender-bias","title":"Men Also Like Shopping: Reducing Gender Bias Amplification using Corpus-level Constraints","arxiv_id":"1707.09457","date":"2017-07-29","proceeding":"EMNLP 2017 9","authors":["Jieyu Zhao","Tianlu Wang","Mark Yatskar","Vicente Ordonez","Kai-Wei Chang"],"abstract":"Language is increasingly being used to define rich visual recognition\nproblems with supporting image collections sourced from the web. Structured\nprediction models are used in these tasks to take advantage of correlations\nbetween co-occurring labels and visual input but risk inadvertently encoding\nsocial biases found in web corpora. In this work, we study data and models\nassociated with multilabel object classification and visual semantic role\nlabeling. We find that (a) datasets for these tasks contain significant gender\nbias and (b) models trained on these datasets further amplify existing bias.\nFor example, the activity cooking is over 33% more likely to involve females\nthan males in a training set, and a trained model further amplifies the\ndisparity to 68% at test time. We propose to inject corpus-level constraints\nfor calibrating existing structured prediction models and design an algorithm\nbased on Lagrangian relaxation for collective inference. Our method results in\nalmost no performance loss for the underlying recognition task but decreases\nthe magnitude of bias amplification by 47.5% and 40.5% for multilabel\nclassification and visual semantic role labeling, respectively.","url_abs":"http://arxiv.org/abs/1707.09457v1","url_pdf":"http://arxiv.org/pdf/1707.09457v1.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":"men-also-like-shopping-reducing-gender-bias","repo_url":"https://github.com/uclanlp/reducingbias","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"men-also-like-shopping-reducing-gender-bias","repo_url":"https://github.com/ARiSE-Lab/DeepInspect","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"men-also-like-shopping-reducing-gender-bias","repo_url":"https://github.com/uclanlp/Fast-and-Robust-Text-Classification","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}}],"tasks":[{"task_slug":"classification","task_name":"General Classification"},{"task_slug":"semantic-role-labeling","task_name":"Semantic Role Labeling"},{"task_slug":"structured-prediction","task_name":"Structured Prediction"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1707.09457","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}