def get_newsgroups():
"""Retrieves the 20 newsgroups dataset"""
datadir = '/local/jlund3/data/'
news_glob = datadir + 'newsgroups-dedup/*/*'
engl_stop = datadir + 'stopwords/english.txt'
news_stop = datadir + 'stopwords/newsgroups.txt'
name_stop = datadir + 'stopwords/malenames.txt'
dataset = ankura.read_glob(news_glob,
tokenizer=ankura.tokenize.news,
labeler=ankura.label.title_dirname)
dataset = ankura.filter_stopwords(dataset, engl_stop)
dataset = ankura.filter_stopwords(dataset, news_stop)
dataset = ankura.combine_words(dataset, name_stop, '<name>')
dataset = ankura.filter_rarewords(dataset, 100)
dataset = ankura.filter_commonwords(dataset, 1500)
for doc, title in enumerate(dataset.titles):
title = title[len(datadir + 'newsgroups-dedup/'):]
dataset.titles[doc] = title
outpath = os.path.join(datadir, 'newsgroups-processed', title)
try:
os.makedirs(os.path.dirname(outpath))
except FileExistsError:
pass
with open(outpath, 'w') as outfile:
tokens = [dataset.vocab[v] for v in dataset.doc_tokens(doc)]
print(' '.join(tokens), file=outfile)
return dataset
def get_newsgroups():
"""Retrieves the 20 newsgroups dataset"""
datadir = '/local/jlund3/data/'
news_glob = datadir + 'newsgroups-dedup/*/*'
engl_stop = datadir + 'stopwords/english.txt'
news_stop = datadir + 'stopwords/newsgroups.txt'
name_stop = datadir + 'stopwords/malenames.txt'
dataset = ankura.read_glob(news_glob,
tokenizer=ankura.tokenize.news,
labeler=ankura.label.title_dirname)
dataset = ankura.filter_stopwords(dataset, engl_stop)
dataset = ankura.filter_stopwords(dataset, news_stop)
dataset = ankura.combine_words(dataset, name_stop, '<name>')
dataset = ankura.filter_rarewords(dataset, 100)
dataset = ankura.filter_commonwords(dataset, 1500)
for doc, title in enumerate(dataset.titles):
title = title[len(datadir + 'newsgroups-dedup/'):]
dataset.titles[doc] = title
outpath = os.path.join(datadir, 'newsgroups-processed', title)
try:
os.makedirs(os.path.dirname(outpath))
except FileExistsError:
pass
with open(outpath, 'w') as outfile:
tokens = [dataset.vocab[v] for v in dataset.doc_tokens(doc)]
print(' '.join(tokens), file=outfile)
return dataset
def get_amazon():
"""Retrieves the amazon dataset"""
text_path = os.path.join(args.data_prefix, 'amazon', 'amazon.txt')
engl_stop = os.path.join(args.data_prefix, 'stopwords/english.txt')
curse_stop = os.path.join(args.data_prefix, 'stopwords/profanity.txt')
dataset = ankura.read_file(text_path, labeler=label.text)
dataset = ankura.filter_stopwords(dataset, engl_stop)
dataset = ankura.combine_words(dataset, curse_stop, '<profanity>')
dataset = ankura.filter_rarewords(dataset, 150)
dataset = ankura.filter_commonwords(dataset, 3000)
dataset = find_display_candidates(dataset, curse_stop)
return dataset
def get_amazon():
"""Retrieves the amazon dataset"""
text_path = os.path.join(args.data_prefix, 'amazon', 'amazon.txt')
engl_stop = os.path.join(args.data_prefix, 'stopwords/english.txt')
curse_stop = os.path.join(args.data_prefix, 'stopwords/profanity.txt')
dataset = ankura.read_file(text_path, labeler=label.text)
dataset = ankura.filter_stopwords(dataset, engl_stop)
dataset = ankura.combine_words(dataset, curse_stop, '<profanity>')
dataset = ankura.filter_rarewords(dataset, 150)
dataset = ankura.filter_commonwords(dataset, 3000)
dataset = find_display_candidates(dataset, curse_stop)
return dataset
def get_newsgroups():
"""Retrieves the 20 newsgroups dataset"""
datadir = '/local/jlund3/data/'
news_glob = datadir + 'newsgroups-dedup/*/*'
engl_stop = datadir + 'stopwords/english.txt'
news_stop = datadir + 'stopwords/newsgroups.txt'
name_stop = datadir + 'stopwords/malenames.txt'
dataset = ankura.read_glob(news_glob,
tokenizer=ankura.tokenize.news,
labeler=ankura.label.title_dirname)
dataset = ankura.filter_stopwords(dataset, engl_stop)
dataset = ankura.filter_stopwords(dataset, news_stop)
dataset = ankura.combine_words(dataset, name_stop, '<name>')
dataset = ankura.filter_rarewords(dataset, 100)
dataset = ankura.filter_commonwords(dataset, 1500)
return dataset
def get_newsgroups():
"""Retrieves the 20 newsgroups dataset"""
news_glob = os.path.join(args.data_prefix, 'newsgroups/*/*')
engl_stop = os.path.join(args.data_prefix, 'stopwords/english.txt')
news_stop = os.path.join(args.data_prefix, 'stopwords/newsgroups.txt')
name_stop = os.path.join(args.data_prefix, 'stopwords/malenames.txt')
curse_stop = os.path.join(args.data_prefix, 'stopwords/profanity.txt')
news_text = functools.partial(label.text, formatter=label.news_formatter)
dataset = ankura.read_glob(news_glob, tokenizer=ankura.tokenize.news,
labeler=news_text)
dataset = ankura.filter_stopwords(dataset, engl_stop)
dataset = ankura.filter_stopwords(dataset, news_stop)
dataset = ankura.combine_words(dataset, name_stop, '<name>')
dataset = ankura.combine_words(dataset, curse_stop, '<profanity>')
dataset = ankura.filter_rarewords(dataset, 100)
dataset = ankura.filter_commonwords(dataset, 1500)
dataset = find_display_candidates(dataset, curse_stop)
return dataset
def get_newsgroups():
"""Retrieves the 20 newsgroups dataset"""
news_glob = os.path.join(args.data_prefix, 'newsgroups/*/*')
engl_stop = os.path.join(args.data_prefix, 'stopwords/english.txt')
news_stop = os.path.join(args.data_prefix, 'stopwords/newsgroups.txt')
name_stop = os.path.join(args.data_prefix, 'stopwords/malenames.txt')
curse_stop = os.path.join(args.data_prefix, 'stopwords/profanity.txt')
news_text = functools.partial(label.text, formatter=label.news_formatter)
dataset = ankura.read_glob(news_glob,
tokenizer=ankura.tokenize.news,
labeler=news_text)
dataset = ankura.filter_stopwords(dataset, engl_stop)
dataset = ankura.filter_stopwords(dataset, news_stop)
dataset = ankura.combine_words(dataset, name_stop, '<name>')
dataset = ankura.combine_words(dataset, curse_stop, '<profanity>')
dataset = ankura.filter_rarewords(dataset, 100)
dataset = ankura.filter_commonwords(dataset, 1500)
dataset = find_display_candidates(dataset, curse_stop)
return dataset
