Polish dataset

To recreate the German dataset, follow these steps:

• Download TIGER treebank files (requires accepting a licence):
  • treebank XML (tigercorpus-2.2.xml.tar.gz),
  • metadata (TIGER2.2.doc.zip),
  • dependency version (tigercorpus-2.2.conll09.tar.gz).
• Convert XML to JSON with this script.
• Produce headed constituencies using this notebook.
• Generate a dataset using these scripts.