added sentiment analysis with running compound score

sentiment_analysis.py

@@ -0,0 +1,128 @@
+from tweepy import Stream, OAuthHandler, StreamListener
+import json
+from datetime import *
+import numpy as np
+from vaderSentiment.vaderSentiment import SentimentIntensityAnalyzer
+from pprint import pprint
+
+timestamp = datetime.strftime(datetime.now(),"%Y%m%d%H%M%S")
+analyzer = SentimentIntensityAnalyzer()
+
+class listener(StreamListener):
+	global i, runningCompound
+	i = 1
+	runningCompound = 0.0
+
+	def on_data(self, data):
+		global i, runningCompound
+
+		try:
+			tweet = json.loads(data)
+			#print len(tweet),tweet['user']['screen_name'],tweet['text']
+			#print tweet['user']['location'],tweet['user']['geo_enabled'],tweet['coordinates'],tweet['geo'],tweet['place']
+
+			if isNotReply("in_reply_to_screen_name", tweet):
+
+				if len(tweets['no_reply'])<N*3:
+					tweets['no_reply'].append(tweet)
+					try:
+						sentence = str(tweet['text'])
+						print(sentence)
+						score = sentimentScore(sentence)
+						#print(score)
+						score = score.get('compound')
+						#print(score)
+						runningCompound = (runningCompound*(i-1) + score)/i
+						i = i + 1
+
+					except Exception as inst:
+						print("Unable to print", inst.args)
+
+				else:
+					timestamp=datetime.strftime(datetime.now(),"%Y%m%d%H%M%S")
+					filename = path2+"dataset_noreply_%s.json"%timestamp
+					f=open(filename,'w')
+					json.dump(tweets['no_reply'],f, sort_keys=True, indent=4)
+					f.close()
+					tweets['no_reply'][:]=[]
+					print (N, "tweets", timestamp)
+					print("\n\n\n", "Running Compound Score: ", runningCompound, "\n\n\n")
+				#print len(tweets['no_reply']),tweet['user']['screen_name'],tweet['text']
+
+			else:
+
+				if len(tweets['in_reply'])<N:
+					tweets['in_reply'].append(tweet)
+
+					try:
+						sentence = str(tweet['text'])
+						print(sentence)
+						# score = sentimentScore(sentence)
+						# #print(score)
+						# score = score.get('compound')
+						# #print(score)
+						# runningCompound = (runningCompound*(i-1) + score)/i
+						# i = i + 1
+
+					except Exception as inst:
+						print("Unable to print", inst.args)
+
+
+					names['replyN'].append(tweet["in_reply_to_screen_name"])
+					names['screenN'].append(tweet['user']['screen_name'])
+				else:
+					timestamp=datetime.strftime(datetime.now(),"%Y%m%d%H%M%S")
+
+					f2=open(path1+"dataset_edges_%s.json"%timestamp,'w')
+
+					for i in np.arange(len(names['replyN'])):
+						f2.write(names['screenN'][i] + "," + names['replyN'][i] + '\n')
+
+					f2.close()
+
+					f=open(path1+"dataset_reply_%s.json"%timestamp,'w')
+					json.dump(tweets['in_reply'],f)
+					f.close()
+					tweets['in_reply'][:]=[]
+					print (N, "tweets", timestamp)
+				#print 'Y', len(tweets['in_reply']),tweet['user']['screen_name'],tweet['text']
+				#print len(tweet["in_reply_to_screen_name"]),tweet["in_reply_to_screen_name"]
+				#sprint
+
+		except Exception as inst:
+			print ("something happened", inst.args)
+
+	def on_error(self,status):
+		if status == 420:
+			# Returning False terminates the Stream to prevent incurring
+			# more errors.
+			return False;
+		print ("Error",status)
+
+
+def isNotReply(string, tweet):
+	return (string not in tweet or tweet[string] == None or tweet[string] == 'null')
+
+def sentimentScore(text):
+	return analyzer.polarity_scores(text)
+
+N=50
+tweets={'in_reply':[],"no_reply":[]}
+names={'replyN':[],"screenN":[]}
+path1="Data1/"
+path2="Data2/"
+ConsumerKey="YpPSDUHum0regAT6vt390zRaR"
+ConsumerSecret="0qpxgfv3SkD6GOZnIAxWVJA7TTewfFoVluPy8Dad5oKigQOxuZ"
+
+AccessToken="2675842215-kXO7G1ZqaVo1hUcA2GUSmBz0Aupnfx5FLaZLvS7"
+AccessTokenSecret="0HpMPy5z1APpsUyQgBBf9qm4eNmcFRFm1k65Nb8EU0CPB"
+
+auth=OAuthHandler(ConsumerKey,ConsumerSecret)
+auth.set_access_token(AccessToken, AccessTokenSecret)
+
+twitterStream=Stream(auth, listener())
+
+#track=["obama","trump"]
+track=["war"]
+[lon0,lat0,lon1,lat1]=[-74.1687,40.5722,-73.8062,40.9467]
+twitterStream.filter(track=track,locations=[lon0,lat0,lon1,lat1])