WMT 2014 French-English and German-English datasets
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/artetxem/monoses
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包括用于训练和评估无监督机器翻译系统的单语语料库。训练数据是通过将2007年至2013年的所有新闻爬虫单语语料库拼接而成。评估则采用Moses工具包中包含的multi-bleu.perl脚本计算的分词BLEU分数。数据集规模分别为：法语7.49亿个标记，德语16.06亿个标记，英语21.09亿个标记。所涉及的任务是机器翻译。
This dataset comprises monolingual corpora for training and evaluating unsupervised machine translation systems. The training data is constructed by concatenating all news-crawled monolingual corpora spanning from 2007 to 2013. For evaluation, the tokenized BLEU score is calculated using the multi-bleu.perl script included in the Moses toolkit. The dataset sizes are as follows: 749 million tokens for French, 1.606 billion tokens for German, and 2.109 billion tokens for English. The task involved is machine translation.
提供机构：
WMT
