def tokenize(self, text: str):
tokens = []
if self.with_postag:
response = self.kytea.getTagsToString(text)
# FIXME Following dirty workaround is required to
# process inputs which include <whitespace> itself
# (e.g. "私 は猫")
response = response.replace("\\ ",
"<SPACE>").replace(" ", " <SPACE>")
for elem in response.split(" ")[:-1]:
# FIXME If input contains a character "/",
# KyTea outputs "//補助記号/・",
# which breaks the simple logic elem.split("/")
pron, postag, surface = map(lambda e: e[::-1],
elem[::-1].split("/", maxsplit=2))
surface = surface.replace("<SPACE>", " ")
tokens.append(Token(surface=surface, postag=postag, pron=pron))
else:
for surface in list(self.kytea.getWS(text)):
tokens.append(Token(surface=surface))
return tokens
def tokenize(self, text: str) -> List[Token]:
return_result = []
parse_result = self.janome.tokenize(text)
if self.with_postag:
for morph in parse_result:
surface = morph.surface
postag, postag2, postag3, postag4 = morph.part_of_speech.split(",")
inflection = morph.infl_type
conjugation = morph.infl_form
base_form = morph.base_form
yomi = morph.reading
pron = morph.phonetic
token = Token(
surface=surface,
postag=postag,
postag2=postag2,
postag3=postag3,
postag4=postag4,
inflection=inflection,
conjugation=conjugation,
base_form=base_form,
yomi=yomi,
pron=pron)
return_result.append(token)
else:
for morph in parse_result:
return_result.append(Token(surface=morph.surface))
return return_result
def tokenize(self, text: str):
"""Tokenize."""
result = []
for token in self.tokenizer.tokenize(text, self.mode):
surface = token.surface()
if self.with_postag:
postag, postag2, postag3, postag4, \
inflection, conjugation = token.part_of_speech()
base_form = token.dictionary_form()
normalized_form = token.normalized_form()
yomi = token.reading_form()
result.append(Token(
surface=surface,
postag=postag,
postag2=postag2,
postag3=postag3,
postag4=postag4,
inflection=inflection,
conjugation=conjugation,
base_form=base_form,
normalized_form=normalized_form,
yomi=yomi,
))
else:
result.append(Token(
surface=surface
))
return result
def test_postagging_with_mecab():
"""Test MeCab tokenizer."""
try:
tokenizer = WordTokenizer(tokenizer="mecab", with_postag=True)
except ImportError:
pytest.skip("natto-py is not installed.")
expect = [Token(**kwargs) for kwargs in mecab_tokens_list]
result = tokenizer.tokenize("吾輩は猫である")
assert expect == result
def test_postagging_with_kytea():
"""Test KyTea tokenizer."""
try:
tokenizer = WordTokenizer(tokenizer="kytea", with_postag=True)
except ImportError:
pytest.skip("KyTea is not installed.")
expect = [Token(**kwargs) for kwargs in kytea_tokens_list]
result = tokenizer.tokenize("吾輩は猫である")
assert expect == result
def test_postagging_with_janome():
"""Test MeCab tokenizer."""
try:
tokenizer = WordTokenizer(tokenizer="janome", with_postag=True)
except ImportError:
pytest.skip("Janome is not installed.")
expect = [Token(**kwargs) for kwargs in janome_tokens_list]
result = tokenizer.tokenize("すもももももももものうち")
assert expect == result
def test_word_tokenize_with_janome():
"""Test MeCab tokenizer."""
try:
tokenizer = WordTokenizer(tokenizer="janome", with_postag=True)
except ImportError:
pytest.skip("skip janome")
expect = [Token(**kwargs) for kwargs in janome_tokens_list]
result = tokenizer.tokenize(SENTENCE3)
assert expect == result
def test_word_tokenize_with_kytea():
"""Test KyTea tokenizer."""
try:
tokenizer = WordTokenizer(tokenizer="kytea", with_postag=True)
except ImportError:
pytest.skip("skip kytea")
expect = [Token(**kwargs) for kwargs in kytea_tokens_list]
result = tokenizer.tokenize(SENTENCE1)
assert expect == result
def test_word_tokenize_with_mecab():
"""Test MeCab tokenizer."""
try:
tokenizer = WordTokenizer(tokenizer="mecab", with_postag=True)
except ImportError:
pytest.skip("skip mecab")
expect = [Token(**kwargs) for kwargs in mecab_tokens_list]
result = tokenizer.tokenize(SENTENCE1)
assert expect == result
def test_word_tokenize_with_kytea():
try:
import Mykytea
del Mykytea
except ImportError:
pytest.skip("Mykytea is not installed.")
tokenizer = WordTokenizer(tokenizer="KyTea")
expect = [Token(surface=w) for w in "吾輩 は 猫 で あ る".split(" ")]
result = tokenizer.tokenize("吾輩は猫である")
assert expect == result
def test_word_tokenize_with_janome():
try:
import janome
del janome
except ImportError:
pytest.skip("janome is not installed")
tokenizer = WordTokenizer(tokenizer="Janome")
expect = [Token(surface=w) for w in "吾輩 は 猫 で ある".split(" ")]
result = tokenizer.tokenize("吾輩は猫である")
assert expect == result
def test_word_tokenize_with_mecab_whitespace():
try:
import natto
del natto
except ImportError:
pytest.skip("natto-py is not installed")
tokenizer = WordTokenizer(tokenizer="MeCab")
expect = [Token(surface=w) for w in "吾輩 は で ある".split(" ")]
result = tokenizer.tokenize("吾輩は である")
assert expect == result
def test_word_tokenize_with_kytea_using_custom_model():
try:
import Mykytea
del Mykytea
except ImportError:
pytest.skip("KyTea is not installed.")
tokenizer = WordTokenizer(tokenizer="KyTea", model_path="data/model.knm")
expect = [Token(surface=w) for w in "吾輩は 猫である".split(" ")]
result = tokenizer.tokenize("吾輩は猫である")
assert expect == result
def test_word_tokenize_with_sudachi_mode_a():
try:
import sudachipy
del sudachipy
except ImportError:
pytest.skip("sudachipy is not installed")
tokenizer = WordTokenizer(tokenizer="Sudachi", mode="A")
expect = [Token(surface=w) for w in "医薬 品 安全 管理 責任 者".split(" ")]
result = tokenizer.tokenize("医薬品安全管理責任者")
assert expect == result
def test_word_tokenize_with_sudachi_mode_a():
"""Test Sudachi tokenizer."""
try:
tokenizer = WordTokenizer(tokenizer="sudachi",
mode="A",
with_postag=True)
except ImportError:
pytest.skip("skip sudachi")
expect = [Token(**kwargs) for kwargs in sudachi_tokens_list]
result = tokenizer.tokenize(SENTENCE2)
assert expect == result
def test_postagging_with_sudachi_mode_a():
"""Test Sudachi tokenizer."""
try:
tokenizer = WordTokenizer(tokenizer="sudachi",
mode="A",
with_postag=True)
except ImportError:
pytest.skip("SudachiPy is not installed.")
expect = [Token(**kwargs) for kwargs in sudachi_tokens_list]
result = tokenizer.tokenize("医薬品安全管理責任者")
assert expect == result
def test_word_tokenize_with_sentencepiece():
try:
import sentencepiece
del sentencepiece
except ImportError:
pytest.skip("Sentencepiece is not installed.")
tokenizer = WordTokenizer(tokenizer="Sentencepiece",
model_path="data/model.spm")
expect = [Token(surface=w) for w in "▁ 吾 輩 は 猫 である".split(" ")]
result = tokenizer.tokenize("吾輩は猫である")
assert expect == result
def test_token_with_postag2():
token = Token(
surface="大崎",
postag="名詞",
postag2="固有名詞,人名,姓",
inflection="*",
conjugation="*",
base_form="大崎",
yomi="オオサキ",
pron="オーサキ")
truth = "名詞,固有名詞,人名,姓,*,*,大崎,オオサキ,オーサキ"
assert token.feature == truth
def tokenize(self, text: str) -> List[Token]:
"""Tokenize"""
return_result = []
parse_result = self.mecab.parse(text).rstrip(" ")
if self.with_postag:
for elem in parse_result.split("\n")[:-1]:
(
surface,
postag,
postag2,
postag3,
postag4,
inflection,
conjugation,
base_form,
yomi,
pron,
) = self.parse_feature(elem)
token = Token(
surface=surface,
postag=postag,
postag2=postag2,
postag3=postag3,
postag4=postag4,
inflection=inflection,
conjugation=conjugation,
base_form=base_form,
yomi=yomi,
pron=pron,
)
return_result.append(token)
else:
for surface in parse_result.split(" "):
return_result.append(Token(surface=surface))
return return_result
def test_kytea_with_s3_model():
try:
import boto3
del boto3
except ImportError:
pytest.skip("skip s3 test because of missing boto3")
try:
tokenizer = WordTokenizer(
tokenizer="KyTea", model_path="s3://konoha-demo/kytea/model.knm")
except ImportError:
pytest.skip("skip kytea")
expect = [Token(surface=w) for w in "吾輩は 猫である".split(" ")] # NOQA
result = tokenizer.tokenize("吾輩は猫である")
assert expect == result
def test_sentencepiece_with_s3_model():
try:
import boto3
del boto3
except ImportError:
pytest.skip("skip s3 test because of missing boto3")
try:
tokenizer = WordTokenizer(
tokenizer="SentencePiece",
model_path="s3://konoha-demo/sentencepiece/model.spm")
except ImportError:
pytest.skip("skip sentencepiece")
expect = [Token(surface=w) for w in "▁ 吾 輩 は 猫 である".split(" ")] # NOQA
result = tokenizer.tokenize("吾輩は猫である")
assert expect == result
def test_word_tokenize_with_whitespace():
tokenizer = WordTokenizer(tokenizer="Whitespace")
expect = [Token(surface=w) for w in "吾輩 は 猫 で ある".split(" ")]
result = tokenizer.tokenize("吾輩 は 猫 で ある")
assert expect == result
def tokenize(self, text: str):
result = []
for subword in self.tokenizer.EncodeAsPieces(text):
token = Token(surface=subword)
result.append(token)
return result
def tokenize(self, text: str):
return [Token(surface=surface) for surface in text.split(" ")]
def test_token_without_feature():
token = Token(surface="大崎")
assert "大崎" == token.surface
assert "" == token.feature
def tokenize(self, text: str):
return [Token(surface=char) for char in list(text)]
def test_word_tokenize_with_character():
tokenizer = WordTokenizer(tokenizer="Character")
expect = [Token(surface=w) for w in "吾 輩 は 猫 で あ る".split(" ")]
result = tokenizer.tokenize("吾輩は猫である")
assert expect == result
def test_token_with_postag():
token = Token(surface="大崎", postag="名詞")
assert "大崎" == token.surface
assert "名詞" == token.feature
