Rule based search

Overview

Camphr provides some rule based matching pipelines: PatternSearcher and RegexRuler, and MultipleRegexRuler. These pipelines are character-based, which means that they are more robust but could be more susceptible to false positives than token-based spaCy pipelines Matcher and PhraseMatcher .

Usage: RegexRuler

1. Create a pipe

   >>> import spacy
   >>> from camphr.pipelines import RegexRuler
   >>> nlp = spacy.blank("en")
   >>> pattern = r"[\d-]+"
   >>> pipe = RegexRuler(pattern, label="PHONE_NUMBER")
   >>> nlp.add_pipe(pipe)
   

2. Parse a text

   >>> text = "My phone number is 012-2345-6666"
   >>> doc = nlp(text)
   >>> print(doc.ents)
   (012-2345-6666,)
   >>> print(doc.ents[0].label_)
   PHONE_NUMBER
   

Usage: MultipleRegexRuler

You can use multiple patterns with MultipleRegexRuler

1. Create a pipe

   >>> import spacy
   >>> from camphr.pipelines import MultipleRegexRuler
   >>> nlp = spacy.blank("en")
   >>> patterns = {"PHONE_NUMBER": r"[\d-]+", "EMAIL": "[\w.]+@[\w.]+"}
   >>> pipe = MultipleRegexRuler(patterns)
   >>> nlp.add_pipe(pipe)
   

2. Parse a text

   >>> text = "Phone: 012-2345-6666, email: bob@foomail.com"
   >>> doc = nlp(text)
   >>> print(doc.ents)
   (012-2345-6666, bob@foomail.com)
   >>> print([e.label_ for e in doc.ents])
   ['PHONE_NUMBER', 'EMAIL']
   

Usage: PatternSearcher

PatternSearcher is useful when you want to look up words based on a large dictionary, thanks to pyahocorasick . This pipeline searches words based on characters, while spaCy provides a similar pipeline PhraseMatcher which is a token-based searcher.

1. Create a pipe

   >>> import spacy
   >>> nlp = spacy.blank("en")
   >>> pipe = PatternSearcher.from_words(["text", "pattern searcher"]) # add words
   >>> nlp.add_pipe(pipe)
   

2. Parse a text

   >>> text = "This is a test text for pattern searcher."
   >>> doc = nlp(text)
   >>> doc.ents
   (text, pattern searcher)