Extraindo frases por padrão de marcações

Single tags are useful, but meaning often lives in groups of words. Tag patterns let you pull out whole phrases instead of lone tokens. 🧩

A noun chunk is a noun plus the words describing it, like the fluffy cat. These chunks usually name the key things a sentence is about.

spaCy finds noun chunks for you. Iterate over doc.noun_chunks to grab each one as a span without writing any rules.

Each noun chunk is a Span, a slice of the Doc. A span knows its text, start, and end, so you can reuse it like any other token group.

To collect actions, keep tokens whose pos_ is VERB. A list comprehension gathers them all in a single readable line.

For your own rules, use the Matcher. You describe a sequence of token attributes, and spaCy finds every place that sequence appears.

A pattern is a list of dictionaries, one per token. This one matches an adjective followed by a noun, like fluffy cat.

Register your pattern under a name with matcher.add(). The name lets you group related rules and identify matches later.

Call the matcher on a Doc to get matches. Each result is a tuple of a match id plus the start and end token positions.

Slice the Doc with the start and end to turn a match into readable text. That slice is the phrase your pattern captured.

Tag patterns understand grammar, so they catch phrases regex would miss. Matching on POS is far cleaner than chasing exact letters.

Think about how to extract grouped nouns.

You extracted phrases two ways: built-in noun_chunks and custom tag patterns with the Matcher. Now you can mine grammar, not just words. 🎉