Any way to import 'ROOT' as a symbol?

[danmysak]

It appears I can check whether a word is the subject of a sentence like this:

nlp('He knows')[0].dep == symbols.nsubj

However,

nlp('He knows')[1].dep == symbols.root

returns False, even though nlp('He knows')[1].dep_ is 'ROOT'. Moreover, nlp('He knows')[1].dep equals to 8206900633647566924, a magic constant (or is it?) that I can’t find anywhere in the source code of the project, except in one of the tests where it is used as a literal.

My question is: is there a way to import the 'ROOT' (or 8206900633647566924) constant from spaCy, or the only option I have is to spell it out as a literal myself?

More generally, what is the reasoning behind asymmetric handling of the relation types such as nsubj or amod (which are included in spacy.symbols) and ROOT or compound (which are not)?

[adrianeboyd]

Which labels are in symbols is just related to historical reasons / backwards compatibility, and in spacy v2+ there's no particular reason to use them directly and I would recommend just referring to the string store. See more details in a related thread here: #8583 (comment)

[danmysak]

Thank you!
I’d like to point out then that the snippet from https://spacy.io/usage/linguistic-features#navigating (the one with from spacy.symbols import nsubj, VERB) is a bit misleading :-)