def test_nltk_punk_sentence_tokenizer_is_used(self):
text = "The wolf killed a duck. What a pitty"
with mock.patch.object(PunktSentenceTokenizer,
'span_tokenize') as nltk_sent:
nltk_sent.return_value = [(0, 5)]
en_tokenize_and_segment(text)
nltk_sent.assert_called_once_with(text)
def test_each_offset_its_the_exac_location_pf_the_token_in_the_text(self):
text = (u"John's bar is cool, right :) XD? "
u"The wolf (starved to death), killed a duck.")
tokens = en_tokenize_and_segment(text)['tokens']
offsets = en_tokenize_and_segment(text)['spans']
for tkn, off in zip(tokens, offsets):
self.assertEqual(text[off:len(tkn) + off], tkn)
def test_sentences_with_big_text(self):
text = (
u"The Bastard Operator From Hell (BOFH), a fictional character "
u"created by Simon Travaglia, is a rogue system administrator who "
u"takes out his anger on users (often referred to as lusers), "
u"colleagues, bosses, and anyone else who pesters him with their "
u"pitiful user created \"problems\".\n"
u"The BOFH stories were originally posted in 1992 to Usenet by "
u"Travaglia, with some being reprinted in Datamation. They were "
u"published weekly from 1995 to 1999 in Network Week and since 2000"
u" they have been published most weeks in The Register. They were "
u"also published in PC Plus magazine for a short time, and several"
u" books of the stories have also been released.")
tokenizer = _get_tokenizer()
expected_sentences = [0]
sentence_splitter = nltk.data.load("tokenizers/punkt/english.pickle")
for i, j in sentence_splitter.span_tokenize(text):
expected_sentences.append(
len(list(tokenizer.span_tokenize(text[:j]))))
sents = en_tokenize_and_segment(text)['sentences']
self.assertEqual(expected_sentences, sents)
def test_there_is_an_offset_per_token(self):
text = u"The wolf (starved to death), killed a duck."
tokens = en_tokenize_and_segment(text)['tokens']
offsets = en_tokenize_and_segment(text)['spans']
self.assertEqual(len(tokens), len(offsets))
def check_expected_words_are_in_tokenization(self, text, expected_words):
words = en_tokenize_and_segment(text)['tokens']
for expected_word in expected_words:
self.assertIn(expected_word, words)
def test_number_of_tokens_is_always_last(self):
text = "The wolf killed a duck. What a pitty"
pieces = en_tokenize_and_segment(text)
sents = pieces['sentences']
tkns = pieces['tokens']
self.assertEqual(sents[-1], len(tkns))
def test_cero_is_all_even_if_no_tokens(self):
text = ""
sents = en_tokenize_and_segment(text)['sentences']
self.assertEqual(sents, [0])
def test_cero_is_always_included(self):
text = "The wolf killed a duck. What a pitty"
sents = en_tokenize_and_segment(text)['sentences']
self.assertEqual(sents[0], 0)