Source code for nltk.tokenize.api

# Natural Language Toolkit: Tokenizer Interface
# Copyright (C) 2001-2021 NLTK Project
# Author: Edward Loper <>
#         Steven Bird <>
# URL: <>
# For license information, see LICENSE.TXT

Tokenizer Interface

from abc import ABC, abstractmethod

from nltk.internals import overridden
from nltk.tokenize.util import string_span_tokenize

[docs]class TokenizerI(ABC): """ A processing interface for tokenizing a string. Subclasses must define ``tokenize()`` or ``tokenize_sents()`` (or both). """
[docs] @abstractmethod def tokenize(self, s): """ Return a tokenized copy of *s*. :rtype: list of str """ if overridden(self.tokenize_sents): return self.tokenize_sents([s])[0]
[docs] def span_tokenize(self, s): """ Identify the tokens using integer offsets ``(start_i, end_i)``, where ``s[start_i:end_i]`` is the corresponding token. :rtype: iter(tuple(int, int)) """ raise NotImplementedError()
[docs] def tokenize_sents(self, strings): """ Apply ``self.tokenize()`` to each element of ``strings``. I.e.: return [self.tokenize(s) for s in strings] :rtype: list(list(str)) """ return [self.tokenize(s) for s in strings]
[docs] def span_tokenize_sents(self, strings): """ Apply ``self.span_tokenize()`` to each element of ``strings``. I.e.: return [self.span_tokenize(s) for s in strings] :rtype: iter(list(tuple(int, int))) """ for s in strings: yield list(self.span_tokenize(s))
[docs]class StringTokenizer(TokenizerI): """A tokenizer that divides a string into substrings by splitting on the specified string (defined in subclasses). """ @property @abstractmethod def _string(self): raise NotImplementedError
[docs] def tokenize(self, s): return s.split(self._string)
[docs] def span_tokenize(self, s): for span in string_span_tokenize(s, self._string): yield span