WMT 2014 English-to-German Translation Task

该数据集包含了用于训练英语到德语机器翻译模型的句子对，采用了Transformer架构。在开发集newstest2013上，通过束搜索方法使用BLEU分数来评估结果。数据规模较大，训练批次中输入和目标句子的token数量大约为4,096个，任务为机器翻译。

This dataset contains sentence pairs for training English-to-German machine translation models that adopt the Transformer architecture. Evaluation is performed using the BLEU score with beam search on the development set newstest2013. The dataset is large-scale, with each training batch comprising approximately 4,096 tokens from both source and target sentences, and its core task is machine translation.

搜集汇总

背景与挑战

背景概述

WMT 2014 English-to-German Translation Task是Tensor2Tensor库中的一个机器翻译数据集，专门用于英语到德语的翻译任务。该数据集支持多种模型配置，特别是Transformer模型，适用于研究和开发高质量的机器翻译系统。