def test_add_node():
graph = Graph()
graph.add_node(Node('a'))
assert Node('a') in graph._adjacency_list.keys()
assert len(graph._adjacency_list.keys()) == 1
assert len(graph._adjacency_list[Node('a')]) == 0
def test_get_nodes():
graph = Graph()
graph.add_node(Node('a'))
graph.add_node(Node('b'))
nodes = graph.get_nodes()
assert len(nodes) == 2
assert Node('a') in nodes and Node('b') in nodes
def test_add_duplicate_edge():
graph = Graph()
graph.add_edge(Node('a'), Node('b'))
graph.add_edge(Node('a'), Node('b'))
assert Node('a') in graph._adjacency_list.keys()
assert Node('b') in graph._adjacency_list.keys()
assert len(graph._adjacency_list.keys()) == 2
assert len(graph._adjacency_list[Node('a')]) == 1
assert len(graph._adjacency_list[Node('b')]) == 0
def test_get_connected_from():
graph = Graph()
graph.add_edge(Node('a'), Node('b'))
graph.add_edge(Node('a'), Node('c'))
assert graph.get_nodes_count() == 3
assert graph.get_edges_count() == 2
connected_from = graph.get_connected_from(Node('a'))
assert len(connected_from) == 2
assert Node('b') in connected_from
assert Node('c') in connected_from
def test_rank():
sentence_list = sorted([
'This is a sentence.', 'A quick brown fox.',
'Jumped over the lazy dog.'
])
sentence_nodes = sentences.rank(
[Node(s, score=sentences.DEFAULT_NODE_SCORE) for s in sentence_list])
sentence_nodes = sorted(sentence_nodes, key=lambda n: n.data)
# Here we expect that each sentence node will have the same score. This is
# because the 'clean' (non-stop) words in each sentence have no similarity
# with any other sentence.
for i in range(len(sentence_list)):
assert sentence_nodes[i].data == sentence_list[i]
assert sentence_nodes[i].score == sentences.DEFAULT_NODE_SCORE
def test_get_nodes_count():
graph = Graph()
graph.add_node(Node('a'))
assert graph.get_nodes_count() == 1
def test_get_averaged_score_data():
node = Node('a', score=10)
assert len(node.variations) == 0
assert node.get_averaged_score() == 10.0
def test_get_averaged_scorevariations():
node = Node('a', score=10)
node.variations = ['v1', 'v2']
assert node.get_averaged_score() == 5.0
def test_equality():
node_a = Node('a')
node_b = Node('a')
assert node_a == node_b
def test_remove_variation():
node = Node('a')
node.variations = ['v1', 'v2']
node.remove_variation('v1')
assert set(node.variations) == set(['v2'])
def test_add_variation():
node = Node('a')
node.add_variation('v')
assert set(node.variations) == set(['v'])
def test_inequality():
node_a = Node('a')
node_b = Node('b')
assert node_a != node_b
def dl_techcrunch(session, stemmer, year, month, day):
base_url = 'https://techcrunch.com/'
archive_url = '{}{}/{}/{}'.format(base_url, year, month, day)
# Download links for the articles.
response = requests.get(archive_url)
if not 200 <= response.status_code < 300:
print('techcrunch failed to download archive page: status {}'.format(
response.status_code))
return False
html = BeautifulSoup(response.text, 'html.parser')
post_titles = html.select('.post-title')
article_urls = []
for post_title in post_titles:
url = post_title.select('a')[0].attrs['href']
if url.replace('www.', '').startswith(base_url):
article_urls.append(url)
# Download, summarise and store articles.
for url in article_urls:
# Check if we already have the article in the DB.
if _get_article(session, normalize_url(url)):
print('~ {}'.format(normalize_url(url)))
continue
tc_article = techcrunch.ArticleLoader.load(url)
# Download the article content.
try:
article = _download_article(url)
# Use for now until summaries and loaders are better.
article.nlp()
except newspaper.article.ArticleException:
print('- {}'.format(normalize_url(url)))
# Normalise and hash all the sentences to make finding the index more
# accurate.
text_sentences = [
_hash_text(s) for s in tokenize_sentences(article.text)]
# Conform data for entry into DB.
summary_sentences = []
keywords = [Node(w) for w in article.keywords]
for tag in tc_article.get('tags', []):
if tag not in article.keywords:
keywords.append(Node(tag))
for sentence in article.summary.split('\n'):
index = text_sentences.index(_hash_text(sentence))
summary_sentences.append(Node(sentence, index=index))
# Insert article and summary into DB.
dao.article.insert(
session=session,
text=tc_article.get('content', article.text),
url=normalize_url(url),
title=tc_article.get('title', article.title),
keywords=keywords,
sentences=summary_sentences,
published_at=_format_timestamp(year, month, day),
s_analysis=None,
)
print('+ {}'.format(url))
return True
def dl_hackernoon(session, stemmer, year, month, day):
base_url = 'https://hackernoon.com/'
archive_url = '{}archive/{}/{}/{}'.format(base_url, year, month, day)
# Download links for the articles.
response = requests.get(archive_url)
if not 200 <= response.status_code < 300:
print('hackernoon failed to download archive page: status {}'.format(
response.status_code))
return False
html = BeautifulSoup(response.text, 'html.parser')
post_title_list = html.select('div.js-postStream')
article_urls = []
if post_title_list:
for anchor in post_title_list[0].select('a'):
url = anchor.attrs['href']
if url.replace('www.', '').startswith(base_url):
if 'source=collection_archive' in url:
url = url.split('?')[0]
match = re.match('[a-z0-9]+', url.split('-')[-1])
if match and url not in article_urls and '@' not in url:
article_urls.append(url)
# Download, summarise and store articles.
for url in article_urls:
# Check if we already have the article in the DB.
if _get_article(session, normalize_url(url)):
print('~ {}'.format(normalize_url(url)))
continue
# Download the article content.
try:
article = _download_article(url)
# Use for now until summaries and loaders are better.
article.nlp()
except newspaper.article.ArticleException:
print('- {}'.format(normalize_url(url)))
hn_article = hackernoon.ArticleLoader.load(url)
# Normalise and hash all the sentences to make finding the index more
# accurate.
text_sentences = [
_hash_text(s) for s in tokenize_sentences(article.text)]
# Conform data for entry into DB.
summary_sentences = []
keywords = [Node(w) for w in article.keywords]
for sentence in article.summary.split('\n'):
index = text_sentences.index(_hash_text(sentence))
summary_sentences.append(Node(sentence, index=index))
# Insert article and summary into DB.
dao.article.insert(
session=session,
text=hn_article.get('content', article.text),
url=normalize_url(url),
title=hn_article.get('title', article.title),
keywords=keywords,
sentences=summary_sentences,
published_at=_format_timestamp(year, month, day),
s_analysis=None,
)
print('+ {}'.format(url))
return True
def has_node(self, data):
'''Return true if there exists a node with the specified data.'''
return Node(data) in self._adjacency_list
def test_get_edges_count():
graph = Graph()
graph.add_edge(Node('a'), Node('b'))
assert graph.get_nodes_count() == 2
assert graph.get_edges_count() == 1
def test_has_node():
graph = Graph()
graph.add_node(Node('a'))
assert graph.has_node('a')
assert not graph.has_node('b')
def test_has_variation():
node = Node('a')
node.variations = ['v']
assert node.has_variation('v')
assert not node.has_variation('w')
def test_get_variations():
node = Node('a')
node.variations = ['v']
assert set(node.get_variations()) == set(['v'])
def _summarize(text='',
title='',
url='',
sentence_count=DEFAULT_SENTENCE_COUNT,
suggestedKeywords=None,
keyword_count=DEFAULT_KEYWORD_COUNT):
article_data = {'title': title, 'text': text, 'url': url}
if url:
# Check if article is cached
article = _get_summary(normalize_url(url))
if article:
if suggestedKeywords is not None:
for word in suggestedKeywords:
newKeyword = Keyword(word, 1)
if newKeyword not in article.keywords:
article.keywords.insert(0, newKeyword)
db.session.add(article)
db.session.commit()
if article.s_analysis is None:
senti_analysis_data = sentiment.SentimentAnalysis.analyise(
article.text)
if senti_analysis_data is not None:
article.s_analysis = senti_analysis_data['label']
return {
'title': article.title,
'text': article.text,
'sentences':
[s.data for s in article.sentences][:sentence_count],
'keywords': [w.data for w in article.keywords][:keyword_count],
's_analysis': article.s_analysis,
}
article = _get_article_from_url(url)
article_data['text'] = article.text
if 'techcrunch' in url:
tc_article = techcrunch.ArticleLoader.load(url)
article_data['title'] = tc_article['title']
article_data['text'] = tc_article['content']
article_data['published_at'] = tc_article['timestamp']
senti_analysis_data = sentiment.SentimentAnalysis.analyise(
tc_article['content'])
print(senti_analysis_data)
if senti_analysis_data is not None:
article_data['s_analysis'] = senti_analysis_data['label']
elif 'wired' in url:
wired_article = wired.ArticleLoader.load(url)
article_data['title'] = wired_article['title']
article_data['text'] = wired_article['content']
article_data['published_at'] = wired_article['date']
senti_analysis_data = sentiment.SentimentAnalysis.analyise(
wired_article['content'])
if senti_analysis_data is not None:
article_data['s_analysis'] = senti_analysis_data['label']
elif 'hackernoon' in url:
hackernoon_article = hackernoon.ArticleLoader.load(url)
article_data['title'] = hackernoon_article['title']
article_data['text'] = hackernoon_article['content']
article_data['published_at'] = hackernoon_article['date']
senti_analysis_data = sentiment.SentimentAnalysis.analyise(
hackernoon_article['content'])
if senti_analysis_data is not None:
article_data['s_analysis'] = senti_analysis_data['label']
elif 'venturebeat' in url:
venturebeat_article = venturebeat.ArticleLoader.load(url)
article_data['title'] = venturebeat_article['title']
article_data['text'] = venturebeat_article['content']
article_data['published_at'] = venturebeat_article['date']
senti_analysis_data = sentiment.SentimentAnalysis.analyise(
venturebeat_article['content'])
if senti_analysis_data is not None:
article_data['s_analysis'] = senti_analysis_data['label']
elif 'news.com.au' in url:
newsau_article = newsau.ArticleLoader.load(url)
article_data['title'] = newsau_article['title']
article_data['text'] = newsau_article['content']
article_data['published_at'] = newsau_article['date']
senti_analysis_data = sentiment.SentimentAnalysis.analyise(
newsau_article['content'])
if senti_analysis_data is not None:
article_data['s_analysis'] = senti_analysis_data['label']
elif 'cricket' in url:
ca_article = cricketau.ArticleLoader.load(url)
article_data['title'] = ca_article['title']
article_data['text'] = ca_article['content']
print(ca_article['content'])
article_data['published_at'] = ca_article['date']
senti_analysis_data = sentiment.SentimentAnalysis.analyise(
ca_article['content'])
if senti_analysis_data is not None:
article_data['s_analysis'] = senti_analysis_data['label']
else:
article_data['title'] = article.title
senti_analysis_data = sentiment.SentimentAnalysis.analyise(
article.text)
if senti_analysis_data is not None:
article_data['s_analysis'] = senti_analysis_data['label']
else:
article_data['title'] = title
senti_analysis_data = sentiment.SentimentAnalysis.analyise(text)
if senti_analysis_data is not None:
article_data['s_analysis'] = senti_analysis_data['label']
sentence_nodes = []
sentences = tokenize_sentences(article_data['text'])
for i, data in enumerate(sentences):
sentence_nodes.append(Node(data, index=i))
ranked_sentences = sorted(rank_sentences(sentence_nodes),
key=lambda n: n.score,
reverse=True)
keywords = rank_words(article_data['title'], article_data['text'])
if suggestedKeywords is not None:
for word in suggestedKeywords:
newKeyword = Keyword(word, 1)
if newKeyword not in article.keywords:
article.keywords.insert(0, newKeyword)
keywords.insert(0, newKeyword)
if url:
_insert_summary(title=article_data['title'],
text=article_data['text'],
url=normalize_url(url),
keywords=keywords,
sentences=ranked_sentences,
published_at=article_data.get('published_at'),
s_analysis=article_data['s_analysis'])
return {
'title': article_data['title'],
'text': article_data['text'],
'sentences': [node.data for node in ranked_sentences][:sentence_count],
'keywords': [node.data for node in keywords][:keyword_count],
's_analysis': article_data['s_analysis']
}
