Association for Computational Linguistics. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 9180–9211, Abu Dhabi, United Arab Emirates. “I’m sorry to hear that”: Finding New Biases in Language Models with a Holistic Descriptor Dataset. Anthology ID: 2022.emnlp-main.625 Volume: Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing Month: December Year: 2022 Address: Abu Dhabi, United Arab Emirates Venue: EMNLP SIG: Publisher: Association for Computational Linguistics Note: Pages: 9180–9211 Language: URL: DOI: Bibkey: smith-etal-2022-im Cite (ACL): Eric Michael Smith, Melissa Hall, Melanie Kambadur, Eleonora Presani, and Adina Williams. We will invite additions and amendments to the dataset, which we hope will serve as a basis for more easy-to-use and standardized methods for evaluating bias in NLP models. We demonstrate that HolisticBias is effective at measuring previously undetectable biases in token likelihoods from language models, as well as in an offensiveness classifier. These descriptors combine with a set of bias measurement templates to produce over 450,000 unique sentence prompts, which we use to explore, identify, and reduce novel forms of bias in several generative models. HolisticBias was assembled in a participatory process including experts and community members with lived experience of these terms. In this work, we present a new, more inclusive bias measurement dataset, HolisticBias, which includes nearly 600 descriptor terms across 13 different demographic axes. Many datasets for measuring bias currently exist, but they are restricted in their coverage of demographic axes and are commonly used with preset bias tests that presuppose which types of biases models can exhibit. I am so sorry to hear about-about everything.Abstract As language models grow in popularity, it becomes increasingly important to clearly measure all possible markers of demographic identity in order to avoid perpetuating existing societal harms.Darling, I am so sorry to hear about Nero's incarceration.I am very sorry to hear colleagues propagating ancient myths like "no solution has been found for dealing with nuclear waste".Sorry to hear about your toxaemia patient, Cynthia.I am sorry to hear of your illness, but since you've sacrificed your health for knowledge of my country, I find you very attractive.I am sorry to hear you are so unhappy, Mrs Izquierdo, but I thought it appropriate to use those four and a half minutes to fill in the background to the situation.I am sorry to hear about your brother, but I need you here.Quite clearly, I am sorry to hear this.I am sorry to hear of your current misfortune.Mr President, I am sorry to hear Mrs Klamt's words. I sure am sorry to hear about that old woman.I am very sorry to hear that, but we are harbouring a fugitive.Mr Fatuzzo, I am very sorry to hear that you cannot sleep at night.I am really sorry to hear that you feel that way, Tami.I am sorry to hear that Mr Mandelson is not sensitive to this problem.I know your parents passed away, and I am sorry to hear that.I am sorry to hear that her presence at Court makes your mother, the Duchess, unhappy.Mr President, I am sorry to hear that the Commissioner is ill this evening.I am sorry to hear that but we will respect your wishes.
0 Comments
Leave a Reply. |
Details
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |