def test_kansuuzi2number(self):
tokens = ['1', '万', '円', '台']
expected = ['10000', '円', '台']
self.assertEqual(kansuuzi2number(tokens), expected)
tokens = ['賞金', 'は' '500', '万']
expected = ['賞金', 'は' '500', '万']
self.assertEqual(kansuuzi2number(tokens), expected)
def test_kansuuzi2number():
tokens = ['1', '万', '円', '台']
expected = ['10000', '円', '台']
result = kansuuzi2number(tokens)
assert result == expected
tokens = ['賞金', 'は' '500', '万']
expected = ['賞金', 'は' '500', '万']
result = kansuuzi2number(tokens)
assert result == expected
def update_headlines(session: Session, user_dict: Path,
logger: Logger) -> None:
query_result = session \
.query(Headline) \
.filter(Headline.is_used.is_(None)) \
.all()
headlines = list(query_result)
if len(headlines) == 0:
return
tokenizer = Tokenizer(str(user_dict))
mappings = []
logger.info('start updating headlines')
for headline in tqdm(headlines):
h = simplify_headline(headline.headline)
is_about_di = headline.categories is not None and \
DOMESTIC_INDEX in headline.categories
# We stopped using `is_template` because the size of the dataset decreased and the result got worse.
# if is_template(h) or not is_interesting(h) or not is_about_di:
if not is_interesting(h) or not is_about_di:
mappings.append({
'article_id': headline.article_id,
'is_used': False
})
continue
tokens = kansuuzi2number(
[token.surface for token in tokenizer.tokenize(h)])
tag_tokens = replace_prices_with_tags(tokens)
mappings.append({
'article_id': headline.article_id,
'simple_headline': h,
'tokens': tokens,
'tag_tokens': tag_tokens,
'is_used': True,
})
session.bulk_update_mappings(Headline, mappings)
session.commit()
logger.info('end updating headlines')