def __init__(self,
tokens=None,
from_file=None,
corpus_type='char',
min_frequency=1,
max_pad_length=None,
sos_eos_tokens=True):
"""Initialization method.
Args:
tokens (list): A list of tokens.
from_file (str): An input file to load the sentences.
corpus_type (str): The desired type to tokenize the sentences. Should be `char` or `word`.
min_frequency (int): Minimum frequency of individual tokens.
max_pad_length (int): Maximum length to pad the tokens.
sos_eos_tokens (bool): Whether start-of-sentence and end-of-sentence tokens should be used.
"""
logger.info('Overriding class: Corpus -> SentenceCorpus.')
# Overrides its parent class with any custom arguments if needed
super(SentenceCorpus, self).__init__(min_frequency=min_frequency)
# Checks if there are not pre-loaded tokens
if not tokens:
# Loads the sentences from file
sentences = l.load_txt(from_file).splitlines()
# Creates a tokenizer based on desired type
pipe = self._create_tokenizer(corpus_type)
# Retrieve the tokens
self.tokens = [pipe(sentence) for sentence in sentences]
# If there are tokens
else:
# Gathers them to the property
self.tokens = tokens
# Cuts the tokens based on a minimum frequency
self._check_token_frequency()
# Pads the tokens before building the vocabulary
self._pad_token(max_pad_length, sos_eos_tokens)
# Builds the vocabulary based on the tokens
self._build()
# Debugging some important information
logger.debug(
'Sentences: %d | Minimum frequency: %d | Maximum pad length: %s | '
'Use <SOS> and <EOS>: %s | Vocabulary size: %d.', len(self.tokens),
self.min_frequency, max_pad_length, sos_eos_tokens,
len(self.vocab))
logger.info('SentenceCorpus created.')
def __init__(self,
tokens=None,
from_file=None,
corpus_type='char',
min_frequency=1):
"""Initialization method.
Args:
tokens (list): A list of tokens.
from_file (str): An input file to load the text.
corpus_type (str): The desired type to tokenize the text. Should be `char` or `word`.
min_frequency (int): Minimum frequency of individual tokens.
"""
logger.info('Overriding class: Corpus -> TextCorpus.')
super(TextCorpus, self).__init__(min_frequency=min_frequency)
# Checks if there are not pre-loaded tokens
if not tokens:
# Loads the text from file
text = l.load_txt(from_file)
# Creates a tokenizer based on desired type
pipe = self._create_tokenizer(corpus_type)
# Retrieve the tokens
self.tokens = pipe(text)
else:
# Gathers them to the property
self.tokens = tokens
# Cuts the tokens based on a minimum frequency
self._check_token_frequency()
# Builds the vocabulary based on the tokens
self._build()
logger.debug(
'Tokens: %d | Minimum frequency: %d | Vocabulary size: %d.',
len(self.tokens), self.min_frequency, len(self.vocab))
logger.info('TextCorpus created.')
def __init__(self, tokens=None, from_file=None, corpus_type='char'):
"""Initialization method.
Args:
tokens (list): A list of tokens.
from_file (str): An input file to load the text.
corpus_type (str): The desired type to tokenize the text. Should be `char` or `word`.
"""
logger.info('Overriding class: Corpus -> TextCorpus.')
# Overrides its parent class with any custom arguments if needed
super(TextCorpus, self).__init__()
# Checks if there are not pre-loaded tokens
if not tokens:
# Loads the text from file
text = l.load_txt(from_file)
# Creates a tokenizer based on desired type
pipe = self._create_tokenizer(corpus_type)
# Retrieve the tokens
self.tokens = pipe(text)
# If there are tokens
else:
# Gathers them to the property
self.tokens = tokens
# Builds the vocabulary based on the tokens
self._build(self.tokens)
# Debugging some important information
logger.debug('Tokens: %d | Vocabulary Size: %d | Type: %s.',
len(self.tokens), len(self.vocab), corpus_type)
logger.info('TextCorpus created.')
import nalp.utils.loader as l
import nalp.utils.preprocess as p
# Loads an input .txt file
text = l.load_txt('data/text/chapter1_harry.txt')
# Creates a character pre-processing pipeline
char_pipe = p.pipeline(p.lower_case, p.valid_char, p.tokenize_to_char)
# Creates a word pre-processing pipeline
word_pipe = p.pipeline(p.lower_case, p.valid_char, p.tokenize_to_word)
# Applying character pre-processing pipeline to text
chars_tokens = char_pipe(text)
# Applying word pre-processing pipeline to text
words_tokens = word_pipe(text)
# Printing tokenized characters and words
print(chars_tokens)
print(words_tokens)
import nalp.utils.loader as l
import nalp.utils.preprocess as p
# Loads an input .txt file with sentences
sentences = l.load_txt('data/sentence/coco_image_captions.txt').splitlines()
# Creates character and word pre-processing pipelines
char_pipe = p.pipeline(p.lower_case, p.valid_char, p.tokenize_to_char)
word_pipe = p.pipeline(p.lower_case, p.valid_char, p.tokenize_to_word)
# Applying character and word pre-processing pipelines to sentences
chars_tokens = [char_pipe(sentence) for sentence in sentences]
words_tokens = [word_pipe(sentence) for sentence in sentences]
# Printing tokenized characters and words
print(chars_tokens)
print(words_tokens)
