Noun chunks for Romanian and rel.manual troubleshooting

Hi there, I'm using the Romanian spacy model, trying to use rel.manual with a patterns file generated from seeds containing 2 labels (I don't have my spans annotated with these labels yet but I was thinking I'd do it all in one go) but I get an error saying that there's no noun chunk iterator for Romanian.

Is it possible to use rel.manual with Romanian and if so what steps should I take? Thank you in advance for your time.

Hi! Are you setting the --add-nps flag when running rel.manual? If there are no noun chunks available in the given language, that feature won't be available, but you should still be able to use the workflow and add noun phrases manually if you want to (e.g. by enabling span highlighting mode with something like --span-label NP, and then merging all noun chunks as an NP span).

(Btw, Romanian noun chunks would definitley be cool to have in spaCy so if you find general rules based on the POS tags that work well, we'd always appreciate a pull request :blush: You can find an example of how these rules look like in the file of the language that implements them, e.g. here for English: spaCy/ at master · explosion/spaCy · GitHub)

1 Like