Материалы к лекции (презентация)

For the study of languages such as OCS, that are mainly extant in translation, an aligned parallel corpus is invaluable. In the PROIEL corpus, all the translations are automatically aligned with the Greek original at token level, with a success rate of about 97 %. In this lecture we discuss the automatic token aligner and demonstrate how the token alignments can be used in combination with multiple layers of annotation to do sophisticated contrastive work on the translation languages.