spaCy 3.0.6版本下nlp.add_pipe添加EntityRuler组件报错的修复方法咨询

The error you're hitting stems from spaCy 3.x's updated pipeline component API—unlike spaCy 2.x, you can’t directly pass a component instance (like your EntityRuler object) to nlp.add_pipe() without specifying a name, or you should use the registered component name string instead. Here’s how to fix your code properly:

Step-by-Step Fixes #

1. Correctly Add the EntityRuler Component #

You have two valid approaches to integrate EntityRuler in spaCy 3.x:

Approach 1: Use the Registered Component Name (Recommended)

For built-in components like EntityRuler, this is the cleanest method:

import spacy
from spacy.pipeline import EntityRuler
from spacy.matcher import Matcher

# Load the pre-trained model
nlp = spacy.load('en_core_web_sm', parse=True, tag=True, entity=True)

# Add the entity_ruler component to the pipeline (prioritize over built-in NER)
ruler = nlp.add_pipe("entity_ruler", before="ner")

# Define your custom animal entities
animal_terms = ["cat", "dog", "arctic fox"]  # Fixed typo: "artic" → "arctic"
# Create patterns and add to the ruler
patterns = [{"label": "ANIMAL", "pattern": term} for term in animal_terms]
ruler.add_patterns(patterns)

Approach 2: Pass the Ruler Instance with a Custom Name

If you prefer creating the EntityRuler first, specify a unique name when adding it to the pipeline:

import spacy
from spacy.pipeline import EntityRuler
from spacy.matcher import Matcher

nlp = spacy.load('en_core_web_sm', parse=True, tag=True, entity=True)

# Initialize the ruler and add patterns
ruler = EntityRuler(nlp)
animal_terms = ["cat", "dog", "arctic fox"]
for term in animal_terms:
    ruler.add_patterns([{"label": "ANIMAL", "pattern": term}])

# Add the custom ruler to the pipeline
nlp.add_pipe(ruler, name="custom_animal_ruler", before="ner")

2. Adjust the Matcher Pattern #

SpaCy uses uppercase entity labels by convention, so update your matcher pattern to match the ANIMAL label (not lowercase "animal"):

# Initialize the matcher
matcher = Matcher(nlp.vocab)
# Fix the pattern to target uppercase entity type
pattern = [{"lower": "no"}, {"ENT_TYPE": {"REGEX": "ANIMAL", "OP": "+"}}]
matcher.add('negated animal', None, pattern)

3. Full Working Code #

Here’s the complete fixed code with all adjustments:

import spacy
from spacy.pipeline import EntityRuler
from spacy.matcher import Matcher

nlp = spacy.load('en_core_web_sm', parse=True, tag=True, entity=True)

# Add entity ruler with custom animal entities
ruler = nlp.add_pipe("entity_ruler", before="ner")
animal_terms = ["cat", "dog", "arctic fox"]
patterns = [{"label": "ANIMAL", "pattern": term} for term in animal_terms]
ruler.add_patterns(patterns)

# Process the input text
doc = nlp("There is no cat in the house and no arctic fox in the basement")

# Merge entities into single tokens for cleaner matching
with doc.retokenize() as retokenizer:
    for ent in doc.ents:
        retokenizer.merge(doc[ent.start:ent.end])

# Set up and run the matcher
matcher = Matcher(nlp.vocab)
pattern = [{"lower": "no"}, {"ENT_TYPE": {"REGEX": "ANIMAL", "OP": "+"}}]
matcher.add('negated animal', None, pattern)

matches = matcher(doc)
for match_id, start, end in matches:
    span = doc[start:end]
    print(span.text)

Key Notes #

• The before="ner" argument ensures your custom entities are recognized before spaCy’s built-in NER model, avoiding conflicts.
• I fixed the typo "artic fox" → "arctic fox" for accuracy, but you can revert it if that was intentional.
• SpaCy 3.x’s pipeline changes are designed to make component configuration more flexible and reproducible (e.g., via config files).

内容的提问来源于stack exchange，提问作者Learner