swap-uniba/xMMEB-train

xMMEB-train数据集是一个多语言机器翻译数据集，它是基于MMEB-train数据集并通过MADLAD 3B模型进行机器翻译得到的。该数据集包含了四种语言版本：法语、德语、意大利语和西班牙语。每个数据集文件包含原始数据集的前10,000个实例，用于训练和评估多模态嵌入任务。此外，还提供了一个并行打乱的文件，用于平行语料库的训练。xMMEB-train dataset is a multilingual machine translation dataset, derived from the MMEB-train dataset through machine translation using the MADLAD 3B model. The dataset includes four language versions: French, German, Italian, and Spanish. Each dataset file contains the first 10,000 instances of the original dataset, used for training and evaluation of multimodal embedding tasks. Additionally, a parallel shuffled file is provided for parallel corpus training.