Best practice for setting custom stop word list non-globally

[pdhall99]

#10942 gives a recommendation for setting a custom stop word list. We might want to have several pipelines in the same session, some of which have default stop words, some of which have custom. Is there a better way to do this than the following, resetting to the default after loading?

import spacy
from spacy.lang.en import EnglishDefaults

EN_DEFAULT_STOP_WORDS = frozenset(EnglishDefaults.stop_words)

# Load pipeline with custom stop_words
custom_stop_words = set(["hello", "world"])
EnglishDefaults.stop_words = custom_stop_words
nlp_custom = spacy.blank("en")
EnglishDefaults.stop_words = set(EN_DEFAULT_STOP_WORDS)  # Reset to defaults

# Load pipeline with default stop_words
nlp_default = spacy.blank("en")

[shadeMe]

Ideally, you'd create a custom language subclass that uses a CustomEnglishDefaults(BaseDefaults) class with your stop words. You can refer to the English language definition to see how that can be done.

If that's not an option, you can simplify the above code by eliminating the EN_DEFAULT_STOP_WORDS global - The default stopwords can be reloaded from spacy.lang.en.stop_words.STOP_WORDS.