def compute_classifier_domain_info():
patterns = set()
# Gets a sample of links from the website starting from the homepage
already_analyzed = set()
links = self._get_all_internal_links(self._start_url)
for i in range(depth - 1):
if len(links) > 3000:
break
sub_links = set()
cpt = 0
for link in links:
print "link %d/%d [%s]" % (cpt, len(links), link)
cpt += 1
if (len(sub_links) > 3000):
break
if not link in already_analyzed:
sub_links = sub_links.union(self._get_all_internal_links(link))
already_analyzed.add(link)
links = links.union(sub_links)
self.log("Finished. Got %d sublinks on %d levels" % (len(links), depth))
# Identifies links pointing to content and generates patterns
ones, zeros = classifier.testSVM(links)
mean, distMat = distmat.getDistanceMatrix(ones)
content = []
for link in distMat:
if sum(distMat[link])/len(distMat) >= mean:
content += [link]
patterns = patternify.getPatterns(content)
return patterns
import sys import crawl import classifier import patternify """usage: python run.py http://cnn.com""" crawler = crawl.crawler() #get a crawler object urls = crawler.crawl(sys.argv[1], 500) #get URLs, number of URLs to crawl classifier.trainSVM() #train preliminary classifier using the "content" and "notcontent" files ones, zeros = classifier.testSVM(urls) #classify using preliminary classifier patterns = patternify.getPatterns(ones) #get patterns from classified "ones" (content links) classifier.trainSVM(patterns) #train secondary classifier with pattern features ones, zeros = classifier.testSVM(urls, patterns) #classify using secondary classifier
