arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2106.04908cs.CLcs.AI
收藏

基于多语言Transformer模型的性别歧视自动检测

Automatic Sexism Detection with Multilingual Transformer Models

Austrian Institute of Technology GmbH(奥地利技术研究所股份公司);St. Pölten University of Applied Sciences(圣珀尔滕应用科技大学)

Mina Schütz, Jaqueline Boeck, Daria Liakhovets, Djordje Slijepčević, Armin Kirchknopf, Manuel Hecht, Johannes Bogensperger, Sven Schlarb, Alexander Schindler, Matthias Zeppelzauer

摘要:

在过去的几年里,性别歧视已经成为社交网络上日益严重的问题。IberLEF 2021上的第一个关于社交网络中的性别歧视识别(EXIST)的共享任务是自然语言处理(NLP)领域的国际竞赛,旨在通过应用机器学习方法自动识别社交媒体内容中的性别歧视。因此,性别歧视检测被公式化为粗(二元)分类问题和细粒度分类任务,该细粒度分类任务区分多种类型的性别歧视内容(例如,支配、刻板印象和客观化)。本文介绍了AIT_FHSTP团队在EXIST 2021基准测试中对这两项任务的贡献。为了解决这些问题,我们应用了两种多语言的Transformer模型,一种基于多语言的BERT,另一种基于XLM-R。我们的方法使用两种不同的策略来适应Transformers的检测性别歧视的内容:第一,无监督的预训练与额外的数据和第二,监督微调与额外的和增强的数据。对于这两项任务,我们的最佳模型是XLM-R,对EXIST数据和其他数据集进行无监督预训练,并对提供的数据集进行微调。二元分类(任务1)的最佳运行达到了0.7752的宏观F1分数,在基准测试中排名第5;对于多类分类(任务2),我们的最佳提交分数排名第6,宏观F1分数为0.5589。

英文摘要:

Sexism has become an increasingly major problem on social networks during the last years. The first shared task on sEXism Identification in Social neTworks (EXIST) at IberLEF 2021 is an international competition in the field of Natural Language Processing (NLP) with the aim to automatically identify sexism in social media content by applying machine learning methods. Thereby sexism detection is formulated as a coarse (binary) classification problem and a fine-grained classification task that distinguishes multiple types of sexist content (e.g., dominance, stereotyping, and objectification). This paper presents the contribution of the AIT_FHSTP team at the EXIST2021 benchmark for both tasks. To solve the tasks we applied two multilingual transformer models, one based on multilingual BERT and one based on XLM-R. Our approach uses two different strategies to adapt the transformers to the detection of sexist content: first, unsupervised pre-training with additional data and second, supervised fine-tuning with additional and augmented data. For both tasks our best model is XLM-R with unsupervised pre-training on the EXIST data and additional datasets and fine-tuning on the provided dataset. The best run for the binary classification (task 1) achieves a macro F1-score of 0.7752 and scores 5th rank in the benchmark; for the multiclass classification (task 2) our best submission scores 6th rank with a macro F1-score of 0.5589.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑