def load_freq_pages(page_view_counts_filename, language, limit=settings["top_articles"]):
"""
Reads a file with page view counts and retrieves the top-k
most frequent page for the language
"""
#OPT: counts file is parsed separately for each language
logging.info("loading list of articles for language %s" % (language))
logging.debug("load freq pages for %s, limit=%s" % (language,limit))
freq_pages = []
#input_file = codecs.open(page_view_counts_filename, 'r', 'utf-8')
input_file = open(page_view_counts_filename, 'r')
for line in input_file:
#line = line.encode('UTF-8')
line = line.rstrip('\n')
(lang, rank, title, count) = line.split(' ')
if lang==language and len(freq_pages) < limit:
title = unquote(title)
try:
id = wikipydia.query_page_id(title, language=lang)
freq_pages.append(title)
except KeyError:
#logging.debug( 'no page for %s %s' % (title, language))
print 'no page for ', title, language
except IOError:
#logging.debug( 'cannot reach %s %s' % (title, language))
print 'cannot reach ', title, language
except TypeError:
#logging.debug( 'unicode object error for %s %s' % (title, language))
print 'unicode object error for ', title, language
except UnicodeDecodeError:
#logging.debug( 'unicode error for %s %s' % (title, language))
print 'unicode error for ', title, language
input_file.close()
logging.info("# of articles loaded: %s" % (len(freq_pages)))
return freq_pages
def write_csv_file(csv_filename,
sentence_filename,
articles_filename,
articles,
lang,
num_sentences_per_hit,
img_output_dir,
img_url_dir,
fontName='Times New Roman',
target_lang='en'):
"""
Generates a comma seperated value file and associated image files
so that a Mechanical Turk translation HIT can be created.
"""
init_files(sentence_filename, articles_filename)
article_ids = []
target_titles = []
write_lines_to_file(articles_filename, articles)
for article in articles:
print article,
article_id = wikipydia.query_page_id(article, language=lang)
print article_id
try:
sentences = get_sentences_for_article(article, article_id, lang,
sentence_filename)
article_ids.append(article_id)
ll = wikipydia.query_language_links(article, lang)
if target_lang in ll:
target_titles.append(ll[target_lang])
else:
target_titles.append('')
except:
target_titles.append('')
write_lines_to_file(articles_filename + '.ids', article_ids)
write_lines_to_file(articles_filename + '.target_titles', target_titles)
#
# translate all sentences
translations = get_translations(sentence_filename, lang)
#
# generate all images
img_urls = generate_images(sentence_filename, img_output_dir, img_url_dir,
fontName)
#
# filter sentences that are mainly ascii
mask = filter_sentences(sentence_filename, lang)
#
csv_output_file = open(csv_filename, 'w')
header = 'lang_pair'
for i in range(1, num_sentences_per_hit + 1):
header += ',seg_id%s' % str(i)
header += ',tag%s' % str(i)
header += ',seg%s' % str(i)
header += ',img_url%s' % str(i)
header += ',machine_translation%s' % str(i)
#
# load the sentences
sentences = read_lines_from_file(sentence_filename)
seg_ids = read_lines_from_file(sentence_filename + '.seg_ids')
tags = read_lines_from_file(sentence_filename + '.tags')
mask = read_lines_from_file(sentence_filename + '.mask')
#
line = header
counter = 0
for i, sentence in enumerate(sentences):
if (mask[i] == '1'):
if counter % num_sentences_per_hit == 0:
csv_output_file.write(line.encode('UTF-8'))
csv_output_file.write('\n'.encode('UTF-8'))
line = lang + "-" + target_lang
counter += 1
seg_id = seg_ids[i]
tag = tags[i]
sentence = format_for_csv(sentence)
img_url = img_urls[i]
translation = format_for_csv(translations[i])
line += ',%s,%s,%s,%s,%s' % (seg_id, tag, sentence, img_url,
translation)
# if there are an odd number of sentences, then fill out the rest of the fields with a do-not-translate message
if not counter % num_sentences_per_hit == 0:
dnt_url = ",,,," + img_url_dir + "/do-not-translate.png,"
line = dnt_url * (num_sentences_per_hit - counter)
csv_output_file.write(line.encode('UTF-8'))
csv_output_file.write('\n'.encode('UTF-8'))
csv_output_file.close()
def write_csv_file(csv_filename, sentence_filename, articles_filename, articles, lang, num_sentences_per_hit, img_output_dir, img_url_dir, fontName='Times New Roman', target_lang='en'):
"""
Generates a comma seperated value file and associated image files
so that a Mechanical Turk translation HIT can be created.
"""
init_files(sentence_filename, articles_filename)
article_ids = []
target_titles = []
write_lines_to_file(articles_filename, articles)
for article in articles:
print article,
article_id = wikipydia.query_page_id(article, language=lang)
print article_id
try:
sentences = get_sentences_for_article(article, article_id, lang, sentence_filename)
article_ids.append(article_id)
ll = wikipydia.query_language_links(article, lang)
if target_lang in ll:
target_titles.append(ll[target_lang])
else:
target_titles.append('')
except:
target_titles.append('')
write_lines_to_file(articles_filename + '.ids', article_ids)
write_lines_to_file(articles_filename + '.target_titles', target_titles)
#
# translate all sentences
translations = get_translations(sentence_filename, lang)
#
# generate all images
img_urls = generate_images(sentence_filename, img_output_dir, img_url_dir, fontName)
#
# filter sentences that are mainly ascii
mask = filter_sentences(sentence_filename, lang)
#
csv_output_file = open(csv_filename, 'w')
header = 'lang_pair'
for i in range(1, num_sentences_per_hit+1):
header += ',seg_id%s' % str(i)
header += ',tag%s' % str(i)
header += ',seg%s' % str(i)
header += ',img_url%s' % str(i)
header += ',machine_translation%s' % str(i)
#
# load the sentences
sentences = read_lines_from_file(sentence_filename)
seg_ids = read_lines_from_file(sentence_filename + '.seg_ids')
tags = read_lines_from_file(sentence_filename + '.tags')
mask = read_lines_from_file(sentence_filename + '.mask')
#
line = header
counter = 0
for i, sentence in enumerate(sentences):
if(mask[i] == '1'):
if counter % num_sentences_per_hit == 0 :
csv_output_file.write(line.encode('UTF-8'))
csv_output_file.write('\n'.encode('UTF-8'))
line = lang + "-" + target_lang
counter += 1
seg_id = seg_ids[i]
tag = tags[i]
sentence = format_for_csv(sentence)
img_url = img_urls[i]
translation = format_for_csv(translations[i])
line += ',%s,%s,%s,%s,%s' % (seg_id, tag, sentence, img_url, translation)
# if there are an odd number of sentences, then fill out the rest of the fields with a do-not-translate message
if not counter % num_sentences_per_hit == 0:
dnt_url = ",,,," + img_url_dir + "/do-not-translate.png,"
line = dnt_url * (num_sentences_per_hit - counter)
csv_output_file.write(line.encode('UTF-8'))
csv_output_file.write('\n'.encode('UTF-8'))
csv_output_file.close()
