Is there a way to manually set the pos tag of a certain token in advance?

[SilasTHU]

I want to specify some words with manual labeled pos tags. And SpaCy do the rest for me based on that. Is that possible?

Thank you for your help.

[polm]

Yes, you can do that in a straightforward way.

import spacy
nlp = spacy.load("en_core_web_sm")
doc = nlp.make_doc("I like cheese") # just tokenize
doc[0].tag_ = "VBP" # "I" is now a verb
doc2 = nlp(doc) # this predicts everything but doesn't overwrite existing settings

Note this depends on the overwrite flag in the Tagger and also possibly the AttributeRuler. See the docs.

[adrianeboyd]

But note that the tagger component won't take this manual tag into account in its predictions. It will make the same predictions for the surrounding words as if the tag weren't there, so you can't use this approach to change the tagger's output.

[SilasTHU]

Thank you! It's great!!!