Load a pre-trained local udpipe of unsupported language

[betimd]

I want to load a local pretrained udpipe model with code like below:

import spacy_udpipe

udpipe_model = spacy_udpipe.UDPipeModel("sq")
nlp = spacy.load("./models_sq", udpipe_model=udpipe_model)

text = "Wikipedia është një uebsite i mirë."

doc = nlp(text)
for token in doc:
    print(token.text, token.lemma_, token.pos_, token.dep_)

but I get Exception: 'sq' language not available

Is it possible to give me direction how to add Albanian language support (or at least skip the exception) and load local udpipe model?

Thank you for your contribution!

[asajatovic]

It should already be possible - please check out #1 and #2 :)

[betimd]

Thanks for a prompt reply.

Is the code above correct or you suggest to load in any other way?

[asajatovic]

Based on the snippet you provided, the following should work:

import spacy_udpipe

nlp = spacy_udpipe.load_from_path(lang="sq",
                                  path="./models_sq",
                                  meta={"description": "Custom sq model"})

text = "Wikipedia është një uebsite i mirë."
doc = nlp(text)
for token in doc:
    print(token.text, token.lemma_, token.pos_, token.dep_)

meta arg is optional info about the model, if not provided, see what it defaults to here.

[asajatovic]

Closing this issue as code snippet 'solution' is added in README.md in #12.