Intrinsic Evaluation Dataset for Automatic Sentence Alignment

This dataset is for the intrinsic evaluation of automatic sentence alignment. It consists of manually aligned sentence pairs from the publicly available German-French parallel corpus Text+Berg (github.com/rsennrich/Bleualign/tree/master/eval) and two newly built corpora based on the English, German, French, Japanese and Italian translations of the Chinese novel The Three-Body Problem and political documents selected from The Governance of China.

All the sentence pairs in the dataset have been checked by human annotators, which can be used as the gold standard to evaluate the output of automatic sentence aligners.