Creating a custom named entity in spaCy

If spaCy's built-in named entities aren't enough, you can make your own using spaCy's entity_ruler pipeline component.

The entity_ruler component allows you to create your own entities and add them to a spaCy pipeline.

You start by adding the component to the existing pipeline with nlp.add_pipe(), placing it before the "ner" component so your own rules are applied first. add_pipe() returns the component it added.

You can then call add_patterns() on the component and pass it a dictionary of the text pattern you'd like to label with an entity.

Since Acme is a technology company, you decide to tag the pattern "smartphone" with the "PRODUCT" entity tag.

spaCy has been imported and nlp holds a loaded en_core_web_sm model. Since changing a pipeline has no effect on docs that already exist, you'll recreate the doc from the transcribed text of call_4_channel_2.wav file once the new component is in place.

This exercise is part of the course

Spoken Language Processing in Python

View Course

Exercise instructions

Add the "entity_ruler" component to the nlp pipeline, before the "ner" component, and save it to ruler.
Add "smartphone" as the value for the "pattern" key.
Print the entity attributes contained in doc.

Hands-on interactive exercise

Have a go at this exercise by completing this sample code.

# Add an entity_ruler component to the pipeline, before ner
ruler = nlp.add_pipe(____, before=____)

# Define pattern for new entity
ruler.add_patterns([{"label": "PRODUCT", "pattern": ____}])

# Recreate the doc so it uses the updated pipeline
doc = nlp(call_4_channel_2_text)

# Test new entity
for entity in doc.____:
  print(entity.text, entity.label_)

Edit and Run Code