Universal Dependencies (UD) is a project for cross-linguistically consistent treebank annotation,
which gained a great popularity since its first release two years ago.
The newest release containing 70 dependency treebanks for 50 languages is used in the CoNLL 2017 multilingual parsing shared task.
This talk will focus on the Udapi framework (available in Python, Perl and Java)
for processing UD data and on related tools for visualizing, editing and converting dependency treebanks.
Udapi is a successor of the Treex framework developed at ÚFAL, with an emphasis on speed, simplicity and multi-linguality.
In addition to showing a few real-world use cases of Udapi, I would like to
present also selected design decisions and more general thoughts about toolkit development.