How to match on pattern when POS doesn't apprear

[antonpibm]

Hi everyone.

I'm having trouble writing a Matcher rule for validating that a certain POS doesn't appear before some other pattern.
Example: validate that no PROPN exist in a sentence before the first VERB

Currently, to solve this I have to instantiate 2 Matchers:

1. For validating a VERB exist, with pattern - [{"POS": "VERB"}]
2. For validating a PROPN exist somewhere before the VERB, with pattern - [{"POS": "PROPN"},{"OP": "*"},{"POS": "VERB"}]

Then in a filtering I need to check if matcher1 and not matcher2

Trying to negate the PROPN with pattern [{"IS_SENT_START": True, "POS": {"NOT_IN": ["PROPN"]},"OP": "*"}, {"POS": "VERB"}] doesn't work, you can test againt: "I am in Brazil, which is the most beautiful country"

I guess I can check the matches programmatically, but I wonder if this could be handles via the pattern itself

[koaning]

I think the "NOT_IN" syntax should work. But when I test against "I am in Brazil, which is the most beautiful country" then on the en_core_web_md model the token "am" is detected as AUX instead of VERB. That might explain part of what you're experiencing here.

import spacy
from spacy.matcher import Matcher
from spacy.displacy import render

nlp = spacy.load("en_core_web_md")

doc = nlp("I am in Brazil, which is the most beautiful country")
render(doc)

Could you share a code snippet with an assert statement that indicates the expected behavior?

Also, could you perhaps share a bit more about the actual use-case you're trying to solve?