The dataset contains 5462 training samples, 711 validation samples and 725 test samples.
Each sample represents a sentence and includes the following features: sentence ID ('sent_id'),
list of tokens ('tokens'), list of lemmas ('lemmas'), list of UPOS tags ('upos_tags'),
list of Multext-East tags ('xpos_tags), list of morphological features ('feats'),
and list of IOB tags ('iob_tags'), which are encoded as class labels.
