MULTIEURLEX-DOC, WIKI-DOC

MULTIEURLEX-DOC和WIKI-DOC是AWS AI Labs精心策划的两个多语言多标签文档图像分类数据集。MULTIEURLEX-DOC包含23种不同语言的欧盟法律文档，每个文档可能被分配到一个或多个标签，强调文档内容的深度理解。WIKI-DOC则包含非欧洲语言的文档，如中文、日文和阿拉伯文，这些文档不仅包含文本，还有丰富的非文本内容，如表格和图像。这两个数据集的创建旨在推动文档AI模型在多语言环境下的性能评估，特别是在零样本跨语言迁移设置中的应用，以解决现有模型在跨语言泛化能力上的局限性。

MULTIEURLEX-DOC and WIKI-DOC are two multilingual multi-label document image classification datasets carefully curated by AWS AI Labs. MULTIEURLEX-DOC consists of EU legal documents in 23 distinct languages, where each document may be assigned to one or more labels, emphasizing the in-depth understanding of document content. WIKI-DOC, by contrast, contains documents in non-European languages such as Chinese, Japanese and Arabic, which not only include text but also rich non-textual content such as tables and images. These two datasets were developed to advance the performance evaluation of document AI models in multilingual scenarios, particularly for applications under zero-shot cross-lingual transfer settings, so as to address the limitations of existing models in cross-lingual generalization capability.