def getTweets(topic,PATH_TO_RAW_DATA,MAX_TWEETS) : auth=tweepy.OAuthHandler(API_KEY, API_SECRET) auth.set_access_token(ACCESS_TOKEN, ACCESS_TOKEN_SECRET) api = tweepy.API(auth,proxy="http://u.padalkar:[email protected]:3128/") #if using proxy based server,add an argument proxy="http://*****:*****@host:port/" for topic in topics : print topic outFile=PATH_TO_RAW_DATA+'/'+topic+".txt" ctr=0 while ctr < MAX_TWEETS : data=tweepy.Cursor(api.search, q=topic,lang='en').items(500) try: tweets = [data.text.lower().encode('utf-8') for tweet in data ] ctr=ctr+len(tweets) util.listTotxt(outFile,tweets,"a+") except tweepy.error.TweepError : print("Waiting for 15 mins : Rate Limit Restriction "+str(ctr)+" Tweets Extracted\n") time.sleep(60*15) # Rate Limit Restriction on no. of tweets extracted in one 15-min window
def Clustering(outfile,tweets,n_clusters,topic,k) :
np.random.seed(0)
vectorizer = TfidfVectorizer(stop_words='english')
X = vectorizer.fit_transform(tweets)
model = KMeans(n_clusters, init='k-means++', max_iter=500, n_init=20)
model.fit(X)
#Check for empty clusters if any
Tweetclusters=list(model.predict(X))
nonempty=[ i for i in range(0,n_clusters) if not Tweetclusters.count(i)==0 ]
empty=list(set(range(0,n_clusters))-set(nonempty))
print("Empty Clusters :"+str(topic)+" TOP "+str(k)+" KMEANS "+str(n_clusters)+" "+str(empty))
#Write Tweetwise Clusters to File
outfile1=os.path.dirname(outfile)+"/TWEET_CLUSTER_"+str(topic)+"_TOP_"+str(k)+"_KMEANS_"+str(n_clusters)+".txt"
util.listTotxt(outfile1,Tweetclusters,"w+")
#Get top ranked tweets from cluster
ind=int(k)/int(n_clusters)
TopTweet ,ClusterAllIndex ,AllTopTweet = [] , [] , []
for i in range(n_clusters) :
ClusterTweets , ClusterIndex , TopClusterTweet=[] , [] ,[]
for j in range(0,len(Tweetclusters)):
if Tweetclusters[j]==i :
ClusterTweets.append(tweets[j])
ClusterIndex.append(j)
ClusterAllIndex.append(ClusterIndex)
TopClusterTweet=rank_by_val(ClusterTweets,topic,ind)
TopTweet.append(TopClusterTweet)
AllTopTweet.extend(TopClusterTweet)
outfile1=os.path.dirname(outfile)+"/INDEX_"+str(topic)+"_TOP_"+str(k)+"_KMEANS_"+str(n_clusters)+".txt"
util.listTotxt(outfile1,ClusterAllIndex,"w+")
with open (outfile,"w+") as f:
for i in range(len(AllTopTweet)) :
j=AllTopTweet[i]
f.write(str(tweets[j]).encode("utf-8")+"\n")
outfile1=os.path.dirname(outfile)+"/CLUSTER_TWEETS_"+str(topic)+"_TOP_"+str(k)+"_KMEANS_"+str(n_clusters)+".txt"
util.listTocsv(outfile1,TopTweet,"w+")
def process(inPath,outPath,topics) :
for topic in topics :
inFile=inPath+'/'+topic+".csv"
tweets=util.csvTolist(inFile)
tweets= [ str(tweet).strip("[']") for tweet in tweets ]
print("No. of Tweets extracted "+str(topic)+"\t\t\t"+str(len(tweets)))
tweets=make_lowercase(tweets)
tweets=remove_repetition(tweets)
tweets=remove_newline(tweets)
tweets=if_not_topic(tweets,topic.lower())
#POS-Tagging of tweets
pos_tweets=tagger.runtagger_parse(tweets) #[[[tw1_token1,postag,confidence],[tw1_token2,postag,confidence]],[[tw2_token1,postag,confidence]]]
tweets=common_except_url(pos_tweets)
pos_tweets=tagger.runtagger_parse(tweets)
print("No. of Tweets after cleaning :"+str(topic)+"\t\t\t"+str(len(tweets)))
outFile=outPath+'/data_'+topic+".txt"
util.listTotxt(outFile,tweets,"w+")
outFile=outPath+'/POS_'+topic+".csv"
util.listTocsv(outFile,pos_tweets,"w+")
def GreedyNormal(outfile,TweetTokens,aspects,tweets,limit): TopInd , ctr =[] , [] left=[aspect for aspect in aspects] for i in range(0,len(TweetTokens)) : x=[ intersect(TweetTokens[i],left) , i ] x.append(len(x[0])) ctr.append(x) ctr=sorted(ctr,key=lambda x: int(x[2]),reverse=True) while len(left)>0 and len(TopInd)<limit : TweetAsp=ctr[0][0] #aspects for that tweet TopInd.append(ctr[0][1])#index for tweet selected left=[ token for token in left if token not in TweetAsp ] # remove aspect from other tweets aspects too as it is covered ctr.remove(ctr[0]) for i in range(0,len(ctr)) : ctr[i][0]=intersect(ctr[i][0], left) ctr[i][2]=len(ctr[i][0]) ctr=sorted(ctr,key=lambda x: int(x[2]),reverse=True) results=[ tweets[i] for i in TopInd ] util.listTotxt( outfile, results , "w+") return results
def getTweets(topic, PATH_TO_RAW_DATA, MAX_TWEETS):
auth = tweepy.OAuthHandler(API_KEY, API_SECRET)
auth.set_access_token(ACCESS_TOKEN, ACCESS_TOKEN_SECRET)
api = tweepy.API(
auth, proxy="http://*****:*****@HOSTNAME:PORT/"
) #if using proxy based server,add an argument proxy="http://*****:*****@host:port/"
for topic in topics:
print topic
outFile = PATH_TO_RAW_DATA + '/' + topic + ".txt"
ctr = 0
while ctr < MAX_TWEETS:
data = tweepy.Cursor(api.search, q=topic, lang='en').items(500)
try:
tweets = [data.text.lower().encode('utf-8') for tweet in data]
ctr = ctr + len(tweets)
util.listTotxt(outFile, tweets, "a+")
except tweepy.error.TweepError:
print("Waiting for 15 mins : Rate Limit Restriction " +
str(ctr) + " Tweets Extracted\n")
time.sleep(
60 * 15
) # Rate Limit Restriction on no. of tweets extracted in one 15-min window
def ranker(rfile,data,topic,ind) :
TopIndex=rank_by_val(data,topic,ind)
TopData=[ data[j].encode("utf-8") for j in TopIndex ]
util.listTotxt(rfile,TopData,"w+")
return TopData
topk = "TOP_" + str(ki)
CosineSimilarityVSM = []
method = "ALL_TWEETS"
outPath = PATH_TO_RESULTS + "/" + topic + "/" + topk + "/" + method
util.createFilePath(outPath)
val = measures.entropy(tweets)
print(topic + topk + method + " Entropy : " + str(val))
method = "RANDOM_TWEETS"
outPath = PATH_TO_RESULTS + "/" + topic + "/" + topk + "/" + method
util.createFilePath(outPath)
results = tweets[0:ki]
rfile = outPath + "/" + topic + "_" + topk + "_" + method + ".txt"
print rfile
util.listTotxt(rfile, results, "w+")
val = measures.entropy(results)
print(topic + " " + topk + method + " Entropy : " + str(val))
measures.get_ParaphraseSim(tweets, rfile, outPath, topic, ki)
CosineSimilarityVSM.append(measures.get_VSMsim(rfile, tweets, results))
outFile = outPath + "/" + topic + "_" + topk + "_" + method + "_VSMSimilarityMatrix.csv"
measures.writeCosineSimMatrix(outFile, tweets, results)
ind = outPath.rfind("/")
outFile = outPath[
0:
ind] + "/" + topic + "_" + topk + "_" + method + "_Doc2vecSimilarityMatrix.csv"
measures.writeDoc2vecSimMatrix(outFile, tweets, results, True)
# UNSUPERVISED CLUSTERING USING KMEANS
n_clusters = [5, 10, 25, 50, 100]
topk="TOP_"+str(ki)
CosineSimilarityVSM=[]
method="ALL_TWEETS"
outPath=PATH_TO_RESULTS+"/"+topic+"/"+topk+"/"+method
util.createFilePath(outPath)
val=measures.entropy(tweets)
print(topic+topk+method+" Entropy : "+str(val))
method="RANDOM_TWEETS"
outPath=PATH_TO_RESULTS+"/"+topic+"/"+topk+"/"+method
util.createFilePath(outPath)
results=tweets[0:ki]
rfile=outPath+"/"+topic+"_"+topk+"_"+method+".txt"
print rfile
util.listTotxt(rfile,results,"w+")
val=measures.entropy(results)
print(topic+" "+topk+method+" Entropy : "+str(val))
measures.get_ParaphraseSim(tweets,rfile,outPath,topic,ki)
CosineSimilarityVSM.append(measures.get_VSMsim(rfile,tweets,results))
outFile=outPath+"/"+topic+"_"+topk+"_"+method+"_VSMSimilarityMatrix.csv"
measures.writeCosineSimMatrix(outFile,tweets,results)
ind=outPath.rfind("/")
outFile=outPath[0:ind]+"/"+topic+"_"+topk+"_"+method+"_Doc2vecSimilarityMatrix.csv"
measures.writeDoc2vecSimMatrix(outFile,tweets,results,True)
# UNSUPERVISED CLUSTERING USING KMEANS
n_clusters=[5,10,25,50,100]
for n in n_clusters :
if ki>=n :
