Stanza includes UD (Universal Dependencies) models of which Faroese is a part of. These models cover functionalities including tokenization, multi-word token (MWT) expansion, lemmatization, part-of-speech (POS) and morphological features tagging and dependency parsing.
The Faroese model is made using FarPaHC. It demonstrates high accuracy for basic tasks tagging tasks but faces challenges with parsing tasks, likely influenced by the small dataset used to train this model.
