How to remove the default patterns on EntityRuler?

[jmaister]

I am working in a Jupyter Notebook for running EntityRuler:

patterns = [
    {"label": "LABEL_ONE", "pattern": [...]},
    {"label": "LABEL_TWO", "pattern": [...]}
]

nlp = spacy.load('en')
ruler = EntityRuler(nlp, validate=True)
ruler.add_patterns(patterns)

nlp.add_pipe(ruler)

print(ruler.labels) # prints ('DRAW_MORE', 'DRAW_ONE')
print(ruler.patterns) # prints my patterns

    t = """ my text """
    doc = nlp(t)
    labels = [ent.label_ for ent in doc.ents]

When I run the rule and get "doc.ents" it shows other labels like 'CARDINAL', 'ORDINAL', 'GPE', 'ORG', 'PERSON', etc...

How can I remove those rules from EntityRuler? I only want my rules to run on the text.

Info about spaCy

• spaCy version: 2.2.3
• Platform: Linux-4.15.0-76-generic-x86_64-with-debian-buster-sid
• Python version: 3.7.4
• Models: en

[svlandeg]

The labels 'CARDINAL' etc. you're seeing, are not coming from default rules in the EntityRuler, but are in fact produced by a pretrained NER component that is part of the pretrained model you're loading. To remove this component, you can either run your code in a with nlp.disable_pipes("ner"): block to disable it temporarily, or you can remove it completely by running nlp.remove_pipe("ner").

The other option you have is to start with a blank model, e.g. nlp = English() which doesn't have any pretrained components (also no sentence splitter!).

[jmaister]

Thank you @svlandeg . It's working as I expected now.