def __init__(self, a_document_name):
# term counts: number of documents that contain a given term (key)
self.term_counts = dict()
self.reviews = dict()
file = open(a_document_name)
for review in file:
review_id = TextProcess.read_line(review)['review_id'].encode('utf8')
review_content = TextProcess.read_line(review)['text'].encode('utf8')
self.reviews[review_id] = review_content
tokens = TextProcess.stemming(TextProcess.stopword(TextProcess.tokenize(TextProcess.read_line(review))))
unique_tokens = list(set(tokens))
for token in unique_tokens:
if token in self.term_counts:
self.term_counts[token] += 1
else:
self.term_counts[token] = 1
self.idf = dict()
for term, term_count in self.term_counts.iteritems():
self.idf[term] = log(len(self.reviews)/term_count)
self.tfidf_dict = dict()
for review_id, review in self.reviews.iteritems():
tokens = TextProcess.stemming(TextProcess.stopword(re.findall(r'\w+',review.lower())))
tf = collections.Counter(tokens)
review_tfidfs = list()
for term, term_count in self.term_counts.iteritems():
if term in tf:
review_tfidfs.append((1+log(tf[term]))*self.idf[term])
else:
review_tfidfs.append(0)
self.tfidf_dict[review_id] = review_tfidfs
def text_preproc(text):
text = TextProcess.shrink_whitespace(text)
text = TextProcess.tolower(text)
text = TextProcess.remove_html(text)
text = TextProcess.remove_url(text)
text = TextProcess.remove_number(text)
text = TextProcess.remove_punctuation(text)
text = TextProcess.remove_stopword(text)
text = TextProcess.shrink_empty_line(text)
return text
def text_preproc(text):
text = TextProcess.shrinkWhitespace(text)
text = TextProcess.toLower(text)
text = TextProcess.removeHTML(text)
text = TextProcess.removeURL(text)
text = TextProcess.removeNumber(text)
text = TextProcess.removePunctuation(text)
text = TextProcess.removeStopword(text)
text = TextProcess.shrinkEmptyLine(text)
return text
def getRestaurants(): restaurants = [] inFileName='restaurants.json' f=open(inFileName,'r') for i in range(0, 14303): rest = TextProcess.read_line(f.readline()) restaurants.append(rest) return restaurants
def calc_tweet_scores(self):
'''calculate relevancy of tweets
- 20% from favorite count
- 30% from retweet count
- 30% from number of followers_count
- 20% from user's total number of tweets
(these amounts may be adjusted later)'''
#score = 0
with open('test.json','w') as f1:
json.dump(self.tweets,f1)
for word in self.tweets:
#print word
sortedList =[]
count = 0
totalscore = 0
s = self.tweets[word]
with open('test2.json','w') as f1:
json.dump(s,f1)
for t in self.tweets[word]:
# adds up to 100%
#print t
#print stuff
#for t in stuff:
#print "t is:"
#print t['favorite_count']
print count
#t = s[count]
if word in t['text']:
#print word
#print unicode(t['text'])
score = 0
score += math.log(t['favorite_count']+1,2) * 0.25
score += math.log(t['retweet_count']+1,2) * 0.025
score += math.log(t['user']['followers_count']+1,2) * 0.05
#score += math.log(t['user']['statuses_count']+1,2) * 0.05
totalscore += score
#stemming the texts
tokens = TextProcess.tokenize(t['text'])
#list_of_stem_words = TextProcess.stemming(tokens)
text = ' '.join(tokens).strip()
self.scores.append({ 'id': t['id'], 'text':unicode(text,errors='ignore'), 'score' : score, 'geo':t['geo']['coordinates']})
#print self.scores
count+=1
if (totalscore >=0):
sortedList = sorted(self.scores, key = lambda k: k['score'], reverse=True)[0:100]
sortedList2 = sorted(sortedList, key = lambda k: k['geo'][1], reverse=True)
if word not in self.sortedTweets:
self.sortedTweets[word] = sortedList2
else:
self.sortedTweets[word] += sortedList2
self.scores = []
def __init__(self, a_document_name):
# term counts: number of documents that contain a given term (key)
self.term_counts = dict()
self.reviews = dict()
file = open(a_document_name)
for review in file:
review_id = TextProcess.read_line(review)['review_id'].encode(
'utf8')
review_content = TextProcess.read_line(review)['text'].encode(
'utf8')
self.reviews[review_id] = review_content
tokens = TextProcess.stemming(
TextProcess.stopword(
TextProcess.tokenize(TextProcess.read_line(review))))
unique_tokens = list(set(tokens))
for token in unique_tokens:
if token in self.term_counts:
self.term_counts[token] += 1
else:
self.term_counts[token] = 1
self.idf = dict()
for term, term_count in self.term_counts.iteritems():
self.idf[term] = log(len(self.reviews) / term_count)
self.tfidf_dict = dict()
for review_id, review in self.reviews.iteritems():
tokens = TextProcess.stemming(
TextProcess.stopword(re.findall(r'\w+', review.lower())))
tf = collections.Counter(tokens)
review_tfidfs = list()
for term, term_count in self.term_counts.iteritems():
if term in tf:
review_tfidfs.append((1 + log(tf[term])) * self.idf[term])
else:
review_tfidfs.append(0)
self.tfidf_dict[review_id] = review_tfidfs
print "start training"
classifier = NaiveBayesClassifier.train(train_features)
reviewList = []
confList = []
reviewContent = {}
print "starting sentiment"
cutoff = 0
for line in open('../../data/yelp_academic_dataset_review.json', 'r'):
#cutoff+=1
#if cutoff > 10:
# break
review_json = TextProcess.read_line(line)
review_id = review_json['review_id']
chopPoint = review_json['text'][::-1].find(' ', 400)
if chopPoint > 400:
reviewContent[review_id] = review_json['text'][::-1][:chopPoint][::-1]
posScore = classifier.posScore(word_features(TextProcess.tokenize(review_json, chopPoint)))
negScore = classifier.negScore(word_features(TextProcess.tokenize(review_json, chopPoint)))
else:
reviewContent[review_id] = review_json['text']
posScore = classifier.posScore(word_features(TextProcess.tokenize(review_json, len(review_json['text']))))
negScore = classifier.negScore(word_features(TextProcess.tokenize(review_json, len(review_json['text']))))
reviewList.append((review_id,posScore))
confList.append((review_id, abs(posScore-negScore)))
print "done with sentiment"
from __future__ import division
__author__ = "jtgoen"
import json
from utility import TextProcess
business_dict = dict()
review_dict = dict()
user_dict = dict(dict())
for line in open("yelp_academic_dataset_business.json", "r"):
business_json = TextProcess.read_line(line)
business_dict[business_json["business_id"]] = business_json
for line in open("yelp_academic_dataset_review.json", "r"):
review_json = TextProcess.read_line(line)
review_dict[review_json["review_id"]] = review_json
sentiment_dict = json.loads(open("reviewSentimentStars.json").read())
confidence_dict = json.loads(open("reviewConfidence.json").read())
for review in review_dict:
if review_dict[review]["user_id"] not in user_dict:
user_id = review_dict[review]["user_id"]
user_dict[user_id] = dict(
value=abs(review_dict[review]["stars"] - sentiment_dict[review_dict[review]["review_id"]]), reviews=1.0
)
else:
user_dict[review_dict[review]["user_id"]]["value"] += abs(