class TopicScrapingTask:
def __init__(self, topic_scraper, article_scraping_task):
"""
@type topic_scraper: Scraper
@type article_scraping_task: ArticleScrapingTask
"""
self._topic_scraper = topic_scraper
self._article_scraping_task = article_scraping_task
self._logger = Logger(self.__class__.__name__)
def run(self, topic_url):
for parser in self._topic_scraper.scrape(topic_url):
assert isinstance(parser, TopicParser)
self._logger.info('Scraping topic at %s.' % topic_url)
articles = []
for article in parser.get_articles():
try:
if self._article_scraping_task.run(article):
articles.append(article)
else:
self._logger.warn('Could not parse article body at %s', article.url)
except IOError, e:
self._logger.error('Failed scraping article: %s' % e)
continue
return articles
class ArticleScrapingTask:
def __init__(self, scraper, min_word_count_heuristic=100):
"""
@type scraper: Scraper
@type min_word_count_heuristic: int
"""
self._scraper = scraper
self._min_word_count_heuristic = min_word_count_heuristic
self._logger = Logger(self.__class__.__name__)
pass
def run(self, article):
"""
@type article: Article
"""
if article.url:
self._logger.info('Scraping %s.', article.url)
for parser in self._scraper.scrape(article.url):
assert isinstance(parser, ArticleParser)
# The final URL of the article may be different, during scraping, the scraper
# passing the final URL to each constructed Parser.
article.url = parser.url
title = parser.get_title()
publish_date = parser.get_publish_date()
preview_image_url = parser.get_preview_image_url()
body = parser.get_body()
if title:
article.title = title
if publish_date:
article.publish_date = publish_date
if preview_image_url:
article.preview_image_url = preview_image_url
if body and self._is_article_body(body):
article.body = body
elif article.description and self._is_article_body(article.description):
article.body = article.description
else:
break
return True
return False
def _is_article_body(self, body):
return self._count_words(body) > self._min_word_count_heuristic
@staticmethod
def _count_words(s):
return len(s.split())
class KeywordAlgorithm(Algorithm):
name = 'keyword'
def __init__(self):
Algorithm.__init__(self)
self._score_mapper = ScoreMapper()
self._logger = Logger(self.__class__.__name__)
def train(self, articles, states):
self._params.extractor = KeywordFeatureExtractor(finder=KeywordFinder(), text_key=lambda a: a.title)
features = np.array(self._params.extractor.train_extract(articles))
scores = np.array(self._score_mapper.map_batch_score(states))
regression = LinearRegression(fit_intercept=True)
n_features = features.shape[1]
self._logger.info('Feature space uses %d keywords', n_features)
if n_features >= 100:
param_grid = {
'pca__n_components': range(50, n_features, 50)
}
pca = PCA(n_components=100)
pipeline = Pipeline([('pca', pca), ('regression', regression)])
clf = GridSearchCV(pipeline, param_grid, n_jobs=1, verbose=0, cv=3, score_func=top_item_scorer)
else:
clf = regression
self._params.classifier = clf
#self._params.classifier = regression
self._params.classifier.fit(features, scores)
def score(self, articles):
self._logger.info('Feature space uses %d keywords', self._params.extractor.keyword_count())
features = self._params.extractor.extract(articles)
return self._params.classifier.predict(np.array(features))
class WorkerThread(threading.Thread):
def __init__(self, worker_id, task_queue, completed_queue):
"""
@type worker_id: int
@type task_queue: Queue.Queue
@type completed_queue: Queue.Queue
"""
self._name = '%s-%d' % (self.__class__.__name__, worker_id)
threading.Thread.__init__(self, name=self._name)
self._logger = Logger(self._name)
self._id = worker_id
self._task_queue = task_queue
self._completed_queue = completed_queue
self._continue = True
def stop(self):
self._continue = False
def run(self):
while self._continue:
self.work()
self.exit()
def work(self):
raise NotImplementedError
def exit(self):
self._logger.info('Exiting.')
@classmethod
def initializer(cls, *args, **kwargs):
class _WorkerThread(cls):
def __init__(self, worker_id, task_queue, completed_queue):
cls.__init__(self, worker_id, task_queue, completed_queue, *args, **kwargs)
return _WorkerThread
