spaCy/spacy/en.pyx
2014-09-15 06:32:18 +02:00

102 lines
3.5 KiB
Cython

# cython: profile=True
# cython: embedsignature=True
'''Tokenize English text, using a scheme that differs from the Penn Treebank 3
scheme in several important respects:
* Whitespace is added as tokens, except for single spaces. e.g.,
>>> [w.string for w in EN.tokenize(u'\\nHello \\tThere')]
[u'\\n', u'Hello', u' ', u'\\t', u'There']
* Contractions are normalized, e.g.
>>> [w.string for w in EN.tokenize(u"isn't ain't won't he's")]
[u'is', u'not', u'are', u'not', u'will', u'not', u'he', u"__s"]
* Hyphenated words are split, with the hyphen preserved, e.g.:
>>> [w.string for w in EN.tokenize(u'New York-based')]
[u'New', u'York', u'-', u'based']
Other improvements:
* Email addresses, URLs, European-formatted dates and other numeric entities not
found in the PTB are tokenized correctly
* Heuristic handling of word-final periods (PTB expects sentence boundary detection
as a pre-process before tokenization.)
Take care to ensure your training and run-time data is tokenized according to the
same scheme. Tokenization problems are a major cause of poor performance for
NLP tools. If you're using a pre-trained model, the :py:mod:`spacy.ptb3` module
provides a fully Penn Treebank 3-compliant tokenizer.
'''
# TODO
#The script translate_treebank_tokenization can be used to transform a treebank's
#annotation to use one of the spacy tokenization schemes.
from __future__ import unicode_literals
from libc.stdlib cimport malloc, calloc, free
from libc.stdint cimport uint64_t
cimport lang
from spacy.lexeme cimport lexeme_check_flag
from spacy.lexeme cimport lexeme_string_view
from spacy._hashing cimport PointerHash
from spacy import orth
cdef class English(Language):
"""English tokenizer, tightly coupled to lexicon.
Attributes:
name (unicode): The two letter code used by Wikipedia for the language.
lexicon (Lexicon): The lexicon. Exposes the lookup method.
"""
cdef int _split_one(self, Py_UNICODE* characters, size_t length):
if length == 1:
return 1
if characters[0] == "'" and (characters[1] == "s" or characters[1] == "S"):
return 2
cdef int i = 0
# Leading punctuation
if _check_punct(characters, 0, length):
return 1
# Contractions
elif length >= 3 and characters[length - 2] == "'":
c2 = characters[length-1]
if c2 == "s" or c2 == "S":
return length - 2
if length >= 1:
# Split off all trailing punctuation characters
i = 0
while i < length and not _check_punct(characters, i, length):
i += 1
return i
abbreviations = set(['U.S', 'u.s', 'U.N', 'Ms', 'Mr', 'P'])
cdef bint _check_punct(Py_UNICODE* characters, size_t i, size_t length):
cdef unicode char_i = characters[i]
cdef unicode char_i1 = characters[i+1]
# Don't count appostrophes as punct if the next char is a letter
if characters[i] == "'" and i < (length - 1) and char_i1.isalpha():
return i == 0
if characters[i] == "-":
return False
#and i < (length - 1) and characters[i+1] == '-':
#return False
# Don't count commas as punct if the next char is a number
if characters[i] == "," and i < (length - 1) and char_i1.isdigit():
return False
if characters[i] == "." and i < (length - 1):
return False
if characters[i] == "." and characters[:i] in abbreviations:
return False
return not char_i.isalnum()
EN = English('en', [], [])