WMT 2020 Sentence-Level Direct Assessment dataset

The dataset used in the competition for Sentence-Level Direct Assessment shared task is composed of data extracted from Wikipedia for six language pairs, consisting of high-resource languages English-German (En-De) and English-Chinese (En-Zh), medium-resource languages Romanian-English (Ro-En) and Estonian-English (Et-En), and low-resource languages Sinhala-English (Si-En) and Nepalese-English (Ne-En), as well as a Russian-English (Ru-En) dataset which combines articles from Wikipedia and Reddit.