โทเคน ช่วงข้อความ และออบเจ็กต์เอกสาร

spaCy gives you three building blocks: the Doc, the Token, and the Span. Learn these and the whole library opens up.

A Doc is the full processed document. It holds the original text plus every token and all the analysis spaCy produced.

You can grab one item by position because a Doc is indexable. Indexing returns a single Token, just like a Python list.

A Token is one unit of text plus rich metadata: its part of speech, its lemma, and whether it is punctuation.

Tokens carry handy flags like is_stop, is_alpha, and is_punct. They let you filter words without writing your own checks.

Slice a Doc and you get a Span, a contiguous run of tokens. It is perfect for phrases like a person or a place.

A Span is not a copy. It still points back into the Doc, so it knows its position and shares all the original analysis.

Named entities show up as Spans in doc.ents. Each one bundles the words plus a label like PERSON or ORG.

spaCy also offers noun_chunks, base noun phrases served as Spans. They are a quick way to pull out the things a sentence talks about.

Need sentences? Iterate doc.sents. Each sentence comes back as a Span, so segmentation is already handled for you.

Through all of this the original string is preserved in doc.text. Tokens and spans are views, never destructive edits.

What do you get when you slice a Doc with doc[0:3]?

The Doc holds everything, indexing gives one Token, and slicing gives a Span. Entities, noun chunks, and sentences all arrive as Spans. 🧩