def skimthatevaluation(self, path):
arr = {}
f = open(path + "original_urls", "r")
for line in f:
line = line.split(" ")
arr[line[0]] = line[1]
g = SkimThatGroundTruth(arr, path)
for i in arr:
domain = ""
if re.search(r"news\.cnet", arr[i]) != None:
domain = "cnetnews"
if domain == "":
print "error in domain"
return
bh = crawler.BlogHtml(arr[i], domain)
items = util.getitems(bh.title, bh.blogparas)
response = ratingslearner.predictratings(items)
response = json.loads(response)
if response == None or response["status"] == 400:
print "error evaluating... itemratings could not be calculated"
return
items = response["content"]
"""
for i in items:
print i['text'], i['rating'], i['position']
"""
sorted_items = sorted(items, key=lambda k: k["rating"], reverse=True)
sorted_items = sorted_items[:5]
sorted_items = sorted(sorted_items, key=lambda k: k["position"])
sim = g.comparewithskimthatgroundtruth(sorted_items, i)
print "cosine similarity: ", sim
def getrateditems(self):
response = ratingslearner.predictratings(self.items)
response = json.loads(response)
if response == None or response['status'] == 400:
return []
items = response['content']
''' Debugging '''
"""
for i in items:
print i['text'], i['rating'], i['position']
"""
return items
def evaluatethesummary(self, path, recommender=False):
blogsitems = {}
summaries = []
blogs = []
for (root, dir, files) in os.walk(path):
if "items" in root:
for i in files:
fi = open(root + "/" + i, "r")
name = i.split(".")[0]
blogsitems[name] = pickle.load(fi)
if recommender:
blogstweets = {}
for (root, dir, files) in os.walk(path):
if "tweets" in root:
for i in files:
ft = open(root + "/" + i, "r")
name = i.split(".")[0]
blogstweets[name] = pickle.load(ft)
for i in blogsitems:
r = recommendersystem.Recommender(0.5)
r.creatematrix(blogsitems[i], blogstweets[i])
r.calcaggregateratings()
items = r.getrecommenderrateditems()
if items == []:
print "error evaluating... recommender itemratings could not be calculated"
return
else:
sorted_items = sorted(items, key=lambda k: k["rating"], reverse=True)
sorted_items = sorted_items[:5]
sorted_items = sorted(sorted_items, key=lambda k: k["position"])
summaries.append(sorted_items)
blogs.append(items)
else:
for i in blogsitems:
response = ratingslearner.predictratings(blogsitems[i])
response = json.loads(response)
if response == None or response["status"] == 400:
print "error evaluating... itemratings could not be calculated"
return
items = response["content"]
for i in items:
print i["text"], i["rating"], i["position"]
sorted_items = sorted(items, key=lambda k: k["rating"], reverse=True)
sorted_items = sorted_items[:5]
sorted_items = sorted(sorted_items, key=lambda k: k["position"])
summaries.append(sorted_items)
blogs.append(items)
g = GroundTruth(blogs)
for i in range(len(summaries)):
g.comparewithgroundtruth(summaries[i], i)
def creatematrix(self, items, tweets):
response = ratingslearner.predictratings(items)
response = json.loads(response)
if response == None or response['status'] == 400:
return
items1 = response['content']
"""
for i in items1:
print i['text'], i['rating'], i['position']
"""
self.items = items1
response1 = ratingslearner.predicttweetratings(tweets)
response1 = json.loads(response1)
if response1 == None or response1['status'] == 400:
return
tweets1 = response1['content']
"""
for i in tweets1:
print i['text'], i['predictedratings'], i['id']
"""
self.tweets = tweets1
ratingrow = []
for i in items1:
ratingrow.append(i['rating'])
self.utilitymatrix['docrow'] = ratingrow
id_indexer = {}
for i in tweets1:
self.utilitymatrix[i['id']] = i['predictedratings']
id_indexer[i['id']] = i
self.supporttweets = []
self.supportratings = []
l = len(items1)
for i in range(l):
self.supporttweets.append("")
self.supportratings.append(-1234)
for i in self.utilitymatrix:
if i != 'docrow':
for j in range(l):
if self.utilitymatrix[i][j] > self.supportratings[j]:
self.supportratings[j] = self.utilitymatrix[i][j]
self.supporttweets[j] = id_indexer[i]