def train(self):
for trainer in self.trainingSet:
#Tweets is a list of dictionaries, where each dictionary is a tweet. The keys are the different parts of the tweet
tweetList = TwitterQuery.search(trainer, results = 10)
#Dictionary that maps a word to how often it occurs
wordOccurences = {}
for tweet in tweetList:
#Print the tweet, and ask the user to rate if this statement has a positive or negative sentiment
tweet.printTweet();
rated_sentiment = int(raw_input('Tweet Sentiment: -3 to 3: '))
tweet.content = tweet.content.lower()
tweet.content = tweet.content.replace(trainer, '')
word_list = tweet.content.split()
for word in word_list:
if(rated_sentiment < 0):
if(self.negativeWords.has_key(word)):
self.negativeWords[word] = self.negativeWords[word] + abs(rated_sentiment)
else:
self.negativeWords[word] = abs(rated_sentiment)
elif(rated_sentiment > 0):
if(self.positiveWords.has_key(word)):
self.positiveWords[word] = self.positiveWords[word] + abs(rated_sentiment)
else:
self.positiveWords[word] = abs(rated_sentiment)
self.writeFiles()
def analyzeSentiment(self, tweetCount, time_frame = 0):
#First, normalize the positiveWords and negativeWords dictionaries
posCount = 0
for key in self.positiveWords:
posCount += self.positiveWords[key]
negCount = 0
for key in self.negativeWords:
negCount += float(self.negativeWords[key])
#Normalize
for key in self.positiveWords:
self.positiveWords[key] = float(self.positiveWords[key])/posCount
for key in self.negativeWords:
self.negativeWords[key] = float(self.negativeWords[key])/negCount
positiveSentiment = 0
negativeSentiment = 0
candidateSentimentLevels = []
for candidate in self.candidates:
if(time_frame == 0):
tweetList = TwitterQuery.search(candidate, results = tweetCount)
else:
tweetList = TwitterQuery.search(candidate, results = tweetCount, mintime = time_frame[0], maxtime = time_frame[1])
bestTweet_pos = "nothing interesting"
bestTweet_neg = "nothing interesting"
bestTweet_prating = 0
bestTweet_nrating = 0
for tweet in tweetList:
tweet_sentiment = self.tweetSentiment(tweet)
if (tweet_sentiment[0]> bestTweet_prating):
bestTweet_pos = tweet.content
elif (tweet_sentiment[1] < bestTweet_nrating):
bestTweet_neg = tweet.content
positiveSentiment += tweet_sentiment[0]
negativeSentiment += tweet_sentiment[1]
print 'The sentiment for ' + candidate + ' is ' + str(positiveSentiment - negativeSentiment)
candidateSentimentLevels.append(positiveSentiment-negativeSentiment)
return candidateSentimentLevels
#
query = 'santorum'
GenericWordsList = ['a', 'about', 'after', 'all', 'and', 'any', 'an', 'are', 'as', 'at', 'been', 'before',
'be', 'but', 'by', 'can', 'could', 'did', 'down', 'do', 'first', 'for', 'from', 'good', 'great', 'had',
'has', 'have', 'her', 'he', 'him', 'his', 'if', 'into', 'in', 'is', 'its', 'it', 'I', 'know', 'like',
'little', 'made', 'man', 'may', 'men', 'me', 'more', 'Mr', 'much', 'must', 'my', 'not', 'now', 'no', 'of',
'on', 'one', 'only', 'or', 'other', 'our', 'out', 'over', 'said', 'see', 'she', 'should', 'some', 'so',
'such', 'than', 'that', 'the', 'their', 'them', 'then', 'there', 'these', 'they', 'this', 'time', 'to',
'two', 'upon', 'up', 'us', 'very', 'was', 'were', 'we', 'what', 'when', 'which', 'who', 'will', 'with',
'would', 'you', 'your', query]
GenericWords = {}
for GenericWord in GenericWordsList:
GenericWords[GenericWord] = 1
WordOccurrences = {}
for i in range(10):
TweetList = TwitterQuery.search(query,100,i+1)
tweetcount = 0
for tweet in TweetList:
#Filter out non alphanumerics and overly generic words, and strip out connected punctuation
unfiltered_words = tweet.content.lower().split()
words = []
for word in unfiltered_words:
temp = word.strip('[]{},.<>/?!$%^&*()_-=+|\\;:\'\"')
if(temp.isalnum() and not GenericWords.has_key(temp)):
words.append(temp)
#Delete duplicates
for word in words:
for i in range(words.count(word)-1):
words.remove(word)
#count occurrences
for word in words:
# set dates for query
fromdate = '2017-01-01'
todate = '2020-01-31'
#Define search term
search_term = 'anti social behaviour'
["Knife Crime", "Anti Social Behaviour", "Sexual Offence", "Shoplifting", "Robbery"]
year = 2017
formated_tweets = []
# set the loop for running through the years until 2019
while year <= 2019 :
twitter_querier = tq.TwitterApiConstruct()
#loop through users tweets
#list_of_tweets = twitter_querier.get_user_timeLine_new('syptweet', fromdate, todate, 10)
try:
list_of_tweets = twitter_querier.query_tweets('20km', search_term, fromdate, todate, 'Sheffield, United Kingdom', 100)
if list_of_tweets == []:
break
#loop through each tweet
for tweet in list_of_tweets:
#create a time object from the tweet date
datetime_object = datetime.strptime(tweet.formatted_date, '%a %b %d %H:%M:%S %z %Y')
self.name = word
self.CoWords = {}
self.count = 1
def printWord(self):
print "\n" + self.name + ": count= " + str(self.count) + "\nWords Occurring with this word:"
for CoWord, num in self.CoWords.iteritems():
if num / self.count > 0.50:
print CoWord + ": " + str(num)
#
WordOccurrences = {}
TweetList = TwitterQuery.search("santorum", 100)
for tweet in TweetList:
words = tweet.content.lower().split()
for word in words:
if WordOccurrences.has_key(word):
WordOccurrences[word].count += 1
CurrentWord = WordOccurrences[word]
else:
CurrentWord = Word(word)
WordOccurrences[word] = CurrentWord
for OtherWord in words:
if OtherWord != CurrentWord.name:
if CurrentWord.CoWords.has_key(OtherWord):
CurrentWord.CoWords[OtherWord] += 1
else:
