The corpus contains movie subtitles in various languages with a raw text version and, if training data was available, automatically parsed versions with morphosyntactic annotation in Universal Dependencies (UD) style. Building on Levshina's (2016) ParTy corpus of movie subtitles, we extended her collection to cover more languages and added movies with a particularly broad linguistic range of available subtitles. Since the original corpus only contained raw text, we trained language models with the Python package spaCy (v2.2.3) on UD treebanks (v2.5 and later) to add morphosyntactic annotation. The annotated data was saved in conllu format.
