We can use TURNA for dependency parsing, after updating the code in t5_classifier.py.
For dependency parsing, we would need 2 outputs per token: one for head word's ID (i.e. order in the sentence, e.g. 3), another for the dependency relation (e.g. acl). Dependency relations are finite in number, like POS tags.
We can use TURNA for dependency parsing, after updating the code in
t5_classifier.py
.For dependency parsing, we would need 2 outputs per token: one for head word's ID (i.e. order in the sentence, e.g. 3), another for the dependency relation (e.g.
acl
). Dependency relations are finite in number, like POS tags.