spaCy로 엔터티 추출하기

spaCy ships a trained NER component that finds entities for you. You feed it text and read back labeled spans. 🐍

First load a pipeline that includes NER. The small English model en_core_web_sm is a quick, friendly starting point.

Pass a string into the pipeline to get a Doc object. The Doc holds tokens, sentences, and the entities spaCy found.

Every Doc exposes doc.ents, a tuple of detected entities. Loop over it to inspect each one individually.

Each entity has a .text attribute holding the exact words matched. This is the raw span lifted from your sentence.

Use .label_ (with the trailing underscore) to get a readable label string like PERSON, ORG, or GPE.

One small loop prints both the text and label of every entity, turning a sentence into clean structured data.

Entities also expose start_char and end_char, the offsets in the original string. Handy for highlighting matches later.

Forgot what GPE means? Call spacy.explain for a plain-English description of any entity label.

Each entity is a Span, a slice of the Doc. So it behaves like a mini-document you can further inspect token by token.

A quick comprehension gathers every entity into a list of tuples, ready to save, count, or feed into a report.

Which attribute gives the readable entity label?

Load a model, run nlp() on text, then loop doc.ents reading .text and .label_. That is core NER in spaCy. ✅