def remove_numbers(text):
"""Remove numbers from text."""
# If text is empty, return None.
if not text: return None
# If texts is tokenized, merge tokens.
if is_tokenized(text):
was_tokenized = True
normalized_text = merge_tokens(text)
else:
was_tokenized = False
normalized_text = text
# Remove all tailing white spaces.
normalized_text = normalized_text.strip()
# Replace all special characters with spaces.
normalized_text = sub(get_number_pattern(), r' ', normalized_text)
# Then remove multiple adjacent spaces.
normalized_text = sub(' +', ' ', normalized_text)
# Strip text.
normalized_text = normalized_text.strip()
# If text was tokenized, re-tokenize text.
if was_tokenized:
normalized_text = word_tokenize(normalized_text)
# Return normalized text.
return normalized_text
def convert_case(text, to_lower=True):
"""Converts text to defined case."""
# If text is empty, return None.
if not text: return None
# If texts is tokenized, merge tokens.
if is_tokenized(text):
was_tokenized = True
normalized_text = merge_tokens(text)
else:
was_tokenized = False
normalized_text = text
# If to lower, convert to lower case. Else, convert to upper case.
if to_lower:
normalized_text = normalized_text.lower()
else:
normalized_text = normalized_text.upper()
# If text was tokenized, re-tokenize text.
if was_tokenized:
normalized_text = word_tokenize(normalized_text)
# Return normalized text.
return normalized_text
def remove_end_characters(text):
"""Removes end characters from word token list."""
# If text is empty, return None.
if not text: return None
# If text is not tokenize, tokenize text.
if is_tokenized(text):
was_tokenized = True
normalized_text = merge_tokens(text)
else:
was_tokenized = False
normalized_text = text
normalized_text += ' '
# Replace stopwords with spaces.
normalized_text = sub(get_end_characters_pattern(), r' ', normalized_text)
# Then remove multiple adjacent spaces.
normalized_text = sub(' +', ' ', normalized_text)
# Then strip text.
normalized_text = normalized_text.strip()
if normalized_text[-1] == r'.':
normalized_text = normalized_text[:-1]
# If text was tokenized, then re-tokenize.
if was_tokenized:
normalized_text = word_tokenize(normalized_text)
# Return normalized text.
return normalized_text
def expand_abbreviations(text):
"""Expands contractions in text."""
# If text is empty, return None.
if not text: return None
# If texts is tokenized, merge tokens.
if is_tokenized(text):
was_tokenized = True
normalized_text = merge_tokens(text)
else:
was_tokenized = False
normalized_text = text
# If last character is not space, add space.
try:
if normalized_text[-1] != ' ':
normalized_text += ' '
except IndexError:
print(1)
# Creates abbreviations pattern.
abbreviations_pattern = compile('({})'.format(r'\.?\s|'.join(
get_abbreviation_dict().keys())),
flags=IGNORECASE | DOTALL)
def expand_match(abbreviation):
"""Expands matched contraction."""
# Retrieves matched contraction from string.
match = abbreviation.group(0)
# If last character is space, remove space.
if match[-1] == " ":
match = match[:-1]
remove_space = True
else:
remove_space = False
# If last character is dot, remove dot.
if match[-1] == r'.':
match = match[:-1]
# Find expanded contraction in dictionary, based on contraction key.
expanded_contraction = get_abbreviation_dict().get(match.lower())
if not expanded_contraction:
return abbreviation.group(0)
if remove_space:
expanded_contraction += " "
# Add first character to expanded contraction.
return expanded_contraction
# Replaces contractions with expanded contractions in text.
normalized_text = abbreviations_pattern.sub(expand_match, normalized_text)
# Strip text.
normalized_text = normalized_text.strip()
# If text was tokenized, re-tokenize text.
if was_tokenized:
normalized_text = word_tokenize(normalized_text)
# Return expanded text.
return normalized_text
def clean_text(text, wordtokenize=False):
clean_dict = {}
# Replace multiple whitespaces.
clean_dict['replace_whitespaces'] = replace_whitespaces(text)
# Replace multiple stopwords.
clean_dict['replace_multiple_stopwords'] = replace_multiple_stopwords(clean_dict['replace_whitespaces'])
# Replace apostrophes.
clean_dict['replace_apostrophes'] = replace_apostrophes(clean_dict['replace_multiple_stopwords'])
# Expand contractions.
clean_dict['expand_contractions'] = expand_contractions(clean_dict['replace_apostrophes'])
# Remove hyperlinks.
clean_dict['remove_hyperlinks'] = remove_hyperlinks(clean_dict['expand_contractions'])
# Remove special characters.
clean_dict['remove_special_characters'] = remove_special_characters(clean_dict['remove_hyperlinks'])
# Remove numbers.
clean_dict['remove_numbers'] = remove_numbers(clean_dict['remove_special_characters'])
# Convert to lower case.
clean_dict['convert_case'] = convert_case(clean_dict['remove_numbers'])
# Expand abbreviations.
clean_dict['expand_abbreviations'] = expand_abbreviations(clean_dict['convert_case'])
# Tokenize sentences.
temp_sentence = correct_pontuation(clean_dict['expand_abbreviations'])
temp_sentence = replace_whitespaces(temp_sentence)
clean_dict['sentence_tokenize'] = sentence_tokenize(temp_sentence)
# If sentence tokenize is empty, return None.
if not clean_dict['sentence_tokenize']:
return clean_dict, None
else:
# Remove end characters.
clean_dict['remove_end_characters'] = [remove_end_characters(item) for item in clean_dict['sentence_tokenize'] if len(item) > 1]
# Lemmatize words.
clean_dict['lemmatize'] = [lemmatize_text(item) for item in clean_dict['remove_end_characters'] if len(item) > 1]
# Remove stopwords.
clean_dict['remove_stopwords'] = [remove_stopwords(item) for item in clean_dict['lemmatize'] if len(item) > 1]
# If tokenize, tokenize words.
if wordtokenize:
clean_dict['sentence_word_tokenize'] = [word_tokenize(item, 'whitespace') for item in
clean_dict['remove_stopwords'] if len(item) > 1]
# Return dictionary and cleaned text.
non_tokenized_result = convert_tokens_to_string_of_words(clean_dict['sentence_word_tokenize'])
clean_dict['direct_word_tokenize'] = word_tokenize(non_tokenized_result, 'whitespace')
return clean_dict, non_tokenized_result
def lemmatize_text(text):
if is_tokenized(text):
was_tokenized = True
normalized_text = text
else:
normalized_text = word_tokenize(text, 'whitespace')
was_tokenized = False
pos_tagged_text = pos_tag_text(normalized_text)
lemmatized_text = [
WordNetLemmatizer().lemmatize(word, pos_tag) if pos_tag else word
for word, pos_tag in pos_tagged_text
]
if not was_tokenized:
lemmatized_text = merge_tokens(lemmatized_text)
return lemmatized_text
def expand_contractions(text):
"""Expands contractions in text."""
# If text is empty, return None.
if not text: return None
# If texts is tokenized, merge tokens.
if is_tokenized(text):
was_tokenized = True
normalized_text = merge_tokens(text)
else:
was_tokenized = False
normalized_text = text
# Creates contractions pattern.
contractions_pattern = compile('({})'.format('|'.join(
get_contraction_dict().keys())),
flags=IGNORECASE | DOTALL)
def expand_match(contraction):
"""Expands matched contraction."""
# Retrieves matched contraction from string.
match = contraction.group(0)
# Stores first character for case sensitivity.
first_char = match[0]
# Find expanded contraction in dictionary, based on contraction key.
expanded_contraction = get_contraction_dict().get(match)
# If the contraction could not be found, try again with lower case.
if not expanded_contraction:
expanded_contraction = get_contraction_dict().get(match.lower())
# Add first character to expanded contraction.
expanded_contraction = first_char + expanded_contraction[1:]
return expanded_contraction
# Replaces contractions with expanded contractions in text.
normalized_text = contractions_pattern.sub(expand_match, normalized_text)
# Strip text.
normalized_text = normalized_text.strip()
# If text was tokenized, re-tokenize text.
if was_tokenized:
normalized_text = word_tokenize(normalized_text)
# Return expanded text.
return normalized_text
def replace_apostrophes(text):
"""Replaces apostrophe pattern with '."""
# If text is empty, return None.
if not text: return None
# If texts is tokenized, merge tokens.
if is_tokenized(text):
was_tokenized = True
normalized_text = merge_tokens(text)
else:
was_tokenized = False
normalized_text = text
# Replaces apostrophe pattern with '.
normalized_text = sub(get_apostrophe_pattern(), "'", normalized_text)
# Strip text.
normalized_text = normalized_text.strip()
# If was tokenized, re-tokenize text.
if was_tokenized:
normalized_text = word_tokenize(normalized_text)
# Return normalized text.
return normalized_text
def replace_multiple_stopwords(text):
"""Replaces multiple stopwords with single stopwords."""
# If text is empty, return None.
if not text: return None
# If texts is tokenized, merge tokens.
if is_tokenized(text):
was_tokenized = True
normalized_text = merge_tokens(text)
else:
was_tokenized = False
normalized_text = text
# Replaces apostrophe pattern with '.
normalized_text = sub('[.!?]+', _get_single_match, normalized_text)
# Strip text.
normalized_text = normalized_text.strip()
# If was tokenized, re-tokenize text.
if was_tokenized:
normalized_text = word_tokenize(normalized_text)
# Return normalized text.
return normalized_text
def remove_hyperlinks(text):
"""Remove hyperlinks from text."""
# If text is empty, return None.
if not text: return None
# If is tokenized, merge tokens.
if is_tokenized(text):
was_tokenized = True
normalized_text = merge_tokens(text)
else:
was_tokenized = False
normalized_text = text
# Replace hyperlinks with space.
normalized_text = sub(get_hyperlink_pattern(), r' ', normalized_text)
# Then remove multiple adjacent spaces.
normalized_text = sub(' +', ' ', normalized_text)
# Strip text.
normalized_text = normalized_text.strip()
# If text was tokenized, re-tokenize text.
if was_tokenized:
normalized_text = word_tokenize(normalized_text)
# Return normalized text.
return normalized_text
def remove_stopwords(text):
"""Remove stopwords from word token list"""
# If text is empty, return None.
if not text: return None
# If text is not tokenize, tokenize text.
if is_tokenized(text):
was_tokenized = True
normalized_text = text
else:
was_tokenized = False
normalized_text = word_tokenize(text, 'whitespace')
# Create stopwords list.
stop_set = set(stopwords.words('english'))
stop_set.update(['amp'])
# Filter stopwords from text.
normalized_text = [
token for token in normalized_text if token not in stop_set
]
# If text was not tokenize, merge tokens.
if not was_tokenized:
normalized_text = merge_tokens(normalized_text)
# Return normalized text.
return normalized_text