lgessler's picture
Update README.md
7bc67e4
metadata
language: grc
widget:
  - text: εἴησαν δ’ ἄν οὗτοι Κρῆτες

This is a MicroBERT model for Ancient Greek.

  • Its suffix is -mxp, which means that it was pretrained using supervision from masked language modeling, XPOS tagging, and UD dependency parsing.
  • The unlabeled Ancient Greek data was taken from the Diorisis corpus, totaling 9,058,227 tokens.
  • The UD treebank UD_Ancient_Greek-PROEIL, v2.9, totaling 213,999 tokens, was used for labeled data.

Please see the repository and the paper for more details.