Marcando palavras com spaCy

You do not have to label words by hand. spaCy reads a sentence and assigns a part of speech to every token automatically. ⚡

First load a trained model. The small English model en_core_web_sm knows how to tag, parse, and recognize entities right away.

Call nlp() on a string to get a Doc. The Doc holds every token along with the tags spaCy has already worked out for you.

A Doc is iterable, so you can walk through it token by token. Each token is a rich object carrying its text and grammatical info.

Every token has a pos_ attribute holding its universal tag, such as NOUN or VERB. The trailing underscore gives you the readable string.

For more detail, read token.tag_. It gives a precise label like VBZ for a present-tense verb, beyond the broad pos_ category.

Forgot what a code means? Call spacy.explain() with a tag and it returns a plain-English description you can read instantly.

spaCy decides each tag from the whole sentence, not a lookup table. That is why book can be tagged NOUN or VERB depending on how you use it.

Once tags exist, filtering is easy. Keep only tokens whose pos_ equals VERB to collect every action word in the text.

spaCy tags thousands of words per second, so you can run it over large documents without waiting. It is built for real production work.

The same loaded model also handles parsing and entities. Tagging is just one part of spaCy's full pipeline running on each Doc.

Recall how spaCy exposes part-of-speech labels.

You loaded a model, processed text, and read both pos_ and tag_ for each token. spaCy turns raw sentences into tagged words in just a few lines. 🎉