def main():
from gensim import corpora, models, similarities
import csv
import textedit
import time
import collections
import glob
stopfile = open("D:/Lresult/NV_s5/subrev_1000.csv", "r")
stopdata = csv.reader(stopfile)
stoplist = collections.Counter()
for line in stopdata:
stoplist[line[0]] = 1
stopfile.close()
buslist = glob.glob("D:/Lresult/NVbus/*")
print len(buslist)
##make document bow
for bus in buslist:
dfile = open(bus, "r")
ddata = csv.reader(dfile)
dlist = []
busname = ""
for line in ddata:
busname = line[2]
if (line[0] not in stoplist):
te = line[5]
doc = textedit.textedit(te)
dlist.append(doc)
dfile.close()
texts = [[word for word in document.lower().split()]
for document in dlist]
dictionary = corpora.Dictionary(texts)
dictionary.save_as_text("D:/Lresult/NVbusdict/" + busname + ".dict")
def main(pas_):
import collections
import csv
import textedit
import time
pas = str(pas_)
print "mkcorpus_start", time.ctime()
##make documents
dnum = 0
subfile = open(pas + "subrev_1000.csv", "r")
subdata = csv.reader(subfile)
subdata.next()
for line in subdata:
te = line[5]
doc = textedit.textedit(te)
dlist = doc.split()
wlist = collections.Counter()
for t in dlist:
wlist[t] = wlist[t] + 1
wfile = open(pas + "subrevbow/" + line[0] + ".csv", "wb")
wri = csv.writer(wfile)
wri.writerow(["word", "num"])
wri.writerows(wlist.items())
wfile.close()
subfile.close()
def main():
from gensim import corpora, models, similarities
import csv
import textedit
import time
import collections
import glob
stopfile=open("D:/Lresult/NV_s5/subrev_1000.csv","r")
stopdata=csv.reader(stopfile)
stoplist=collections.Counter()
for line in stopdata:
stoplist[line[0]]=1
stopfile.close()
buslist=glob.glob("D:/Lresult/NVbus/*")
print len(buslist)
##make document bow
for bus in buslist:
dfile=open(bus,"r")
ddata=csv.reader(dfile)
dlist=[]
busname=""
for line in ddata:
busname=line[2]
if(line[0] not in stoplist):
te=line[5]
doc=textedit.textedit(te)
dlist.append(doc)
dfile.close()
texts = [[word for word in document.lower().split()] for document in dlist]
dictionary = corpora.Dictionary(texts)
dictionary.save_as_text("D:/Lresult/NVbusdict/"+busname+".dict")
def main(pas_): import collections import csv import textedit import time pas=str(pas_) print "mkcorpus_start",time.ctime() ##make documents dnum=0 subfile=open(pas+"subrev_1000.csv","r") subdata=csv.reader(subfile) subdata.next() for line in subdata: te=line[5] doc=textedit.textedit(te) dlist=doc.split() wlist=collections.Counter() for t in dlist: wlist[t]=wlist[t]+1 wfile=open(pas+"subrevbow/"+line[0]+".csv","wb") wri=csv.writer(wfile) wri.writerow(["word","num"]) wri.writerows(wlist.items()) wfile.close() subfile.close()
def main(model_,bnum_,tnum_,train_):
from gensim import corpora, models, similarities
import csv
import textedit
import time
pas=""
print "mkcorpus_start",time.ctime()
# remove common words and tokenize
#stopfile=open("stopwords_en.csv","r")
#stopdata=csv.reader(stopfile)
#stoplist=[]
#for line in stopdata:
# stoplist.append(line[0])
#stopfile.close()
#print stoplist
##make documents
dlist=[]
dfile=open("notNVreview.csv","r")
#dfile=open("testrev.csv")
ddata=csv.reader(dfile)
dnum=0
for line in ddata:
te=line[0]
doc=textedit.textedit(te)
dlist.append(doc)
dnum=dnum+1
if(dnum%10000==0):
print dnum,
dfile.close()
#print dlist
print "dfile fin",time.ctime()
texts = [[word for word in document.lower().split()] for document in dlist]
print "text fin",time.ctime()
# remove words that appear only once
#all_tokens = sum(texts, [])
#tokens_once = set(word for word in set(all_tokens) if all_tokens.count(word) == 1)
#texts = [[word for word in text if word not in tokens_once] for text in texts]
#print(texts)
#print("texts fin")
dictionary = corpora.Dictionary(texts)
print "dictionary fin",time.ctime(),len(dictionary.token2id)
dictionary.filter_extremes(no_below=10,no_above=0.5)
print "dictionary cut fin",time.ctime(),len(dictionary.token2id)
dictionary.save("nNVreview.dict")
dictionary.save_as_text(pas+"nNVreview_text.dict")
corpus=[dictionary.doc2bow(text) for text in texts]
corpora.MmCorpus.serialize(pas+"nNVreview.mm", corpus)
#print corpus
print "mk_corpus fin",time.ctime()
def main(clus_):
from gensim import corpora, models, similarities
import csv
import textedit
import time
pas = "******"
clus = str(clus_)
print "mkcorpus_start", time.ctime()
# remove common words and tokenize
#stopfile=open("stopwords_en.csv","r")
#stopdata=csv.reader(stopfile)
#stoplist=[]
#for line in stopdata:
# stoplist.append(line[0])
#stopfile.close()
#print stoplist
##make documents
for clus in range(0, 10):
clus = str(clus)
dlist = []
dfile = open(
"D:/Lresult/ks/o4b6t500LDAbus_p/4bEjOyTaDG24SY5TxsaUNQ.csv", "r")
#dfile=open("testrev.csv")
ddata = csv.reader(dfile)
ddata.next()
dnum = 0
for line in ddata:
if (line[6] == clus):
te = line[5]
doc = textedit.textedit(te)
dlist.append(doc)
dfile.close()
print "dfile fin", time.ctime()
texts = [[word for word in document.lower().split()]
for document in dlist]
print "text fin", time.ctime()
# remove words that appear only once
#all_tokens = sum(texts, [])
#tokens_once = set(word for word in set(all_tokens) if all_tokens.count(word) == 1)
#texts = [[word for word in text if word not in tokens_once] for text in texts]
#print(texts)
#print("texts fin")
dictionary = corpora.Dictionary(texts)
print "dictionary fin", time.ctime(), len(dictionary.token2id)
#dictionary.filter_extremes(no_below=10,no_above=0.5)
#print "dictionary cut fin",time.ctime(),len(dictionary.token2id)
#dictionary.save("clus"+clus+".dict")
dictionary.save_as_text(pas + "clus" + clus + "_text.dict")
def main(clus_):
from gensim import corpora, models, similarities
import csv
import textedit
import time
pas="******"
clus=str(clus_)
print "mkcorpus_start",time.ctime()
# remove common words and tokenize
#stopfile=open("stopwords_en.csv","r")
#stopdata=csv.reader(stopfile)
#stoplist=[]
#for line in stopdata:
# stoplist.append(line[0])
#stopfile.close()
#print stoplist
##make documents
for clus in range(0,10):
clus=str(clus)
dlist=[]
dfile=open("D:/Lresult/ks/o4b6t500LDAbus_p/4bEjOyTaDG24SY5TxsaUNQ.csv","r")
#dfile=open("testrev.csv")
ddata=csv.reader(dfile)
ddata.next()
dnum=0
for line in ddata:
if(line[6]==clus):
te=line[5]
doc=textedit.textedit(te)
dlist.append(doc)
dfile.close()
print "dfile fin",time.ctime()
texts = [[word for word in document.lower().split()] for document in dlist]
print "text fin",time.ctime()
# remove words that appear only once
#all_tokens = sum(texts, [])
#tokens_once = set(word for word in set(all_tokens) if all_tokens.count(word) == 1)
#texts = [[word for word in text if word not in tokens_once] for text in texts]
#print(texts)
#print("texts fin")
dictionary = corpora.Dictionary(texts)
print "dictionary fin",time.ctime(),len(dictionary.token2id)
#dictionary.filter_extremes(no_below=10,no_above=0.5)
#print "dictionary cut fin",time.ctime(),len(dictionary.token2id)
#dictionary.save("clus"+clus+".dict")
dictionary.save_as_text(pas+"clus"+clus+"_text.dict")
def main():
from gensim import corpora
import csv
import textedit
import collections
stopfile=open("D:/Lresult/NV_s5/subrev_1000.csv","r")
stopdata=csv.reader(stopfile)
stoplist=collections.Counter()
for line in stopdata:
te=line[5]
dlist=[]
doc=textedit.textedit(te)
dlist.append(doc)
texts = [word for word in document.lower().split()]
dictionary = corpora.Dictionary(texts)
dictionary.save_as_text("D:/Lresult/NV_s5/subrevdict/"+line[0]+".dict")
stopfile.close()
def main():
import csv
import textedit
import time
import collections
import glob
stopfile = open("D:/Lresult/NV_s5/subrev_1000.csv", "r")
stopdata = csv.reader(stopfile)
stoplist = collections.Counter()
for line in stopdata:
stoplist[line[0]] = 1
stopfile.close()
buslist = glob.glob("D:/Lresult/NVbus/*")
print len(buslist)
##make document bow
for bus in buslist:
dfile = open(bus, "r")
ddata = csv.reader(dfile)
busname = ""
wlist = collections.Counter()
for line in ddata:
busname = line[2]
if (line[0] not in stoplist):
te = line[5]
doc = textedit.textedit(te)
dlist = doc.split()
for t in dlist:
wlist[t] = wlist[t] + 1
wfile = open("D:/Lresult/NVbusbow/" + busname + ".csv", "wb")
writer = csv.writer(wfile)
writer.writerow(["word", "num"])
writer.writerows(wlist.items())
wfile.close()
dfile.close()
def main():
import csv
import textedit
import time
import collections
import glob
stopfile=open("D:/Lresult/NV_s5/subrev_1000.csv","r")
stopdata=csv.reader(stopfile)
stoplist=collections.Counter()
for line in stopdata:
stoplist[line[0]]=1
stopfile.close()
buslist=glob.glob("D:/Lresult/NVbus/*")
print len(buslist)
##make document bow
for bus in buslist:
dfile=open(bus,"r")
ddata=csv.reader(dfile)
busname=""
wlist=collections.Counter()
for line in ddata:
busname=line[2]
if(line[0] not in stoplist):
te=line[5]
doc=textedit.textedit(te)
dlist=doc.split()
for t in dlist:
wlist[t]=wlist[t]+1
wfile=open("D:/Lresult/NVbusbow/"+busname+".csv","wb")
writer=csv.writer(wfile)
writer.writerow(["word","num"])
writer.writerows(wlist.items())
wfile.close()
dfile.close()
from gensim import corpora, models, similarities import csv import textedit import time pas="******" print "start",time.ctime() dlist=[] dfile=open(pas+"nNVreview.csv","r") ddata=csv.reader(dfile) dnum=0 for line in ddata: te=line[0] doc=textedit.textedit(te) dlist.append(doc) dnum=dnum+1 if(dnum%100000==0): print dnum, dfile.close() #print dlist print "dfile fin",time.ctime() texts = [[word for word in document.lower().split()] for document in dlist] print "text fin",len(dlist),time.ctime() ##dictionary_load dictionary=corpora.Dictionary.load(pas+"/corpus_pl/nNVreviewpl.dict") ####stopword_load sfile=open(pas+"stopwords/over4word.csv","r") sdata=csv.reader(sfile) slist=[]
def main(cluster,fwords):
import csv
import os
import collections
import textedit
import numpy
busname="4bEjOyTaDG24SY5TxsaUNQ"
pas="******"
pas2="//kaede/PPTShare/masafumi/musc/20151117"
fewords=int(fwords)####feature word num
cnum=int(cluster)###cluster_num
ifile=open(pas+"/ks/busclus/"+busname+".csv","r")
idata=csv.reader(ifile)
idata.next()
cluslist=collections.Counter()
for line in idata:
####revid,sentiment_num,clus
cluslist[line[0],line[1]]=int(line[2])
ifile.close()
ifile=open(pas2+"/ks/bussent/"+busname+".csv","r")
idata=csv.reader(ifile)
wlist=collections.Counter()####word dictionary for each cluster
for num in range(0,cnum):
wlist[num]=collections.Counter()
for line in idata:
doc=textedit.textedit(line[5])
doc=doc.split()
for t in doc:
wlist[cluslist[line[0],line[6]]][t]=wlist[cluslist[line[0],line[6]]][t]+1
ifile.close()
tflist=collections.Counter()
wordlen=[]
wordsum=collections.Counter()
for t in range(0,cnum):
wordlen.append(len(wlist[t]))
wordsum[t]=sum(wlist[t].values())
for w in wlist[t]:
tflist[w]=tflist[w]+1
print wordlen,max(wordlen)
wfile=open(pas+"/ks/busclusword/"+busname+".csv","wb")
writer=csv.writer(wfile)
header=[]
for num in map(str,range(0,cnum)):
header=header+["c"+num+"word","c"+num+"num","c"+num+"tfidf"]
writer.writerow(header)
for num in range(0,max(wordlen)):
wwlist=[]
for t in range(0,10):
if(len(wlist[t])<=num):
wwlist=wwlist+["_",0,0]
else:
tmp=wlist[t].items()
wwlist.append(tmp[num][0])
wwlist.append(tmp[num][1])
wwlist.append(1.0*tmp[num][1]/wordsum[t]*numpy.log(1+10.0/tflist[tmp[num][0]]))
writer.writerow(wwlist)
wfile.close()
def main(model_, bnum_, tnum_, train_, pas_):
from gensim import corpora, models, similarities
import csv
import textedit
import time
import glob
model = str(model_)
bnum = int(bnum_)
tnum = int(tnum_)
train = str(train_)
pas = str(pas_)
print "mkcorpus_start", time.ctime()
#remove subrev
subfile = open(pas + "subrev_1000.csv", "r")
subdata = csv.reader(subfile)
sublist = []
for line in subdata:
sublist.append(line[0])
subfile.close()
##remove stoplist
stopfile = open(pas + "stopwords/over4word.csv", "r")
stopdata = csv.reader(stopfile)
stoplist = []
for line in stopdata:
stoplist.append(line[0])
stopfile.close()
stopset = set(stoplist)
##make documents
st = "D:/Lresult/NVbus/*"
dlist = glob.glob(st)
print len(dlist), "star"
dnum = 0
for bus in dlist:
dfile = open(bus, "r")
ddata = csv.reader(dfile)
dlist = []
for line in ddata:
if (line[0] not in sublist):
te = line[5]
doc = textedit.textedit(te)
dlist.append(doc)
dnum = dnum + 1
if (dnum % 1000 == 0):
print dnum,
dfile.close()
texts = [[word for word in document.lower().split()]
for document in dlist]
#print "text fin",time.ctime(),texts
texts = [[word for word in text if word not in stopset]
for text in texts]
dictionary = corpora.Dictionary(texts)
#print "dictionary fin",time.ctime(),len(dictionary.token2id)
#dictionary.filter_extremes(no_below=bnum)
#print "dictionary cut fin",time.ctime(),len(dictionary.token2id)
#dictionary.save(pas+train+"_o4b"+str(bnum)+"t"+str(tnum)+".dict")
dictionary.save_as_text(pas + "NVbus_o4b6_bow/" + bus[17:-4] + ".tsv")
ddata = csv.reader(dfile)
ddic = collections.Counter()
for line in ddata:
ddic[line[0]] = map(float, line[1:])
dfile.close()
print "dic fin", len(ddic), len(ddic["we"]), time.ctime()
##base data
revfile = open("D:/Lresult/NV_s5/subrev_1000.csv", "r")
revdata = csv.reader(revfile)
revdata.next()
revid = collections.Counter()
revvec = collections.Counter()
for line in revdata:
revid[line[0]] = line[2] ##bus_id
text = textedit.textedit(line[5])
revvec[line[0]] = numpy.array([0] * int(dim))
for num in range(0, len(text)):
if (num not in stoplist):
revvec[line[0]] = revvec[line[0]] + numpy.array(ddic[text[num]])
revfile.close()
print "rev fin", len(revvec), time.ctime()
####comp data
sfile = open("D:/Lresult/NVreview.csv", "r")
sdata = csv.reader(sfile)
sdata.next()
svec = collections.Counter()
lnum = 0
slist = []
for line in sdata:
def main(model_,bnum_,tnum_,train_,pas_): from gensim import corpora, models, similarities import csv import textedit import time import glob model=str(model_) bnum=int(bnum_) tnum=int(tnum_) train=str(train_) pas=str(pas_) print "mkcorpus_start",time.ctime() #remove subrev subfile=open(pas+"subrev_1000.csv","r") subdata=csv.reader(subfile) sublist=[] for line in subdata: sublist.append(line[0]) subfile.close() ##remove stoplist stopfile=open(pas+"stopwords/over4word.csv","r") stopdata=csv.reader(stopfile) stoplist=[] for line in stopdata: stoplist.append(line[0]) stopfile.close() stopset=set(stoplist) ##make documents st="D:/Lresult/NVbus/*" dlist=glob.glob(st) print len(dlist),"star" dnum=0 for bus in dlist: dfile=open(bus,"r") ddata=csv.reader(dfile) dlist=[] for line in ddata: if(line[0] not in sublist): te=line[5] doc=textedit.textedit(te) dlist.append(doc) dnum=dnum+1 if(dnum%1000==0): print dnum, dfile.close() texts = [[word for word in document.lower().split()] for document in dlist] #print "text fin",time.ctime(),texts texts = [[word for word in text if word not in stopset] for text in texts] dictionary = corpora.Dictionary(texts) #print "dictionary fin",time.ctime(),len(dictionary.token2id) #dictionary.filter_extremes(no_below=bnum) #print "dictionary cut fin",time.ctime(),len(dictionary.token2id) #dictionary.save(pas+train+"_o4b"+str(bnum)+"t"+str(tnum)+".dict") dictionary.save_as_text(pas+"NVbus_o4b6_bow/"+bus[17:-4]+".tsv")
#calc topic sim
header=[]
header.append("rev_id")
header.append("bus_id")
for num in range(1,51):
header.append(str(num))
wfile=open("subrevtopic.csv","wb")
writer=csv.reader(wfile)
writer.writerow(header)
ifile=open("subrev.csv","r")
idata=csv.reader(ifile)
idata.next()
for line in idata:
wlist=[]
wlist.append(line[0])
wlist.append(line[1])
doc=textedit.textedit(line[2])
vec_bow = dictionary.doc2bow(doc.lower().split())
vec_lda = lda[vec_bow]
slist=[0]*50
for num in range(0,len(vec_lda)):
slist[vec_lda[num][0]]=vec_lda[num][1]
wlist=wlist+slist
writer.writerow(wlist)
ifile.close()
wfile.close()
def main(model_,bnum_,tnum_,train_,pas_):
from gensim import corpora, models, similarities
import csv
import textedit
import time
model=str(model_)
bnum=int(bnum_)
tnum=int(tnum_)
train=str(train_)
pas=str(pas_)
print "start",time.ctime()
#dictionary = corpora.Dictionary.load(pas+train+"_o4b"+str(bnum)+"t"+str(tnum)+".dict")
#corpus = corpora.MmCorpus(pas+train+"_o4b"+str(bnum)+"t"+str(tnum)+".mm")
dictionary = corpora.Dictionary.load(pas+"over4/corpus/nNVrev_o4b6.dict")
#use LSI
#lsi = models.LsiModel.load(pas+train+"_o4b"+str(bnum)+"t"+str(tnum)+".lsi")
#if(model=="lda"):
# lsi=models.LdaModel.load(pas+train+"_o4b"+str(bnum)+"t"+str(tnum)+".lda")
lsi = models.LsiModel.load(pas+"over4/model/nNVrevo4b6_t300.lsi")
#calc topic sim
header=[]
header.append("rev_id")
header.append("user_id")
header.append("bus_id")
header.append("stars")
header.append("sentnum")
header.append("date")
for num in range(0,int(tnum)):
header.append("t"+str(num).zfill(len(str(tnum))/10))
#wfile=open(pas+train+model+"_o4b"+str(bnum)+"t"+str(tnum)+".csv","wb")
wfile=open(pas+"model/hoge.csv","wb")
writer=csv.writer(wfile)
writer.writerow(header)
"NVreview.csv:[review_id,user_id,business_id,stars,date,texts]"
#test file
ifile=open(pas+"ks/NVrevrawsent.csv","r")
idata=csv.reader(ifile)
idata.next()
k=0
for line in idata:
wlist=[]
wlist.append(line[0])
wlist.append(line[1])
wlist.append(line[2])
wlist.append(line[3])
wlist.append(line[6])###for revraw only
wlist.append(line[4])
doc=textedit.textedit(line[5])
vec_bow = dictionary.doc2bow(doc.lower().split())
vec_lsi = lsi[vec_bow]
slist=[0]*int(tnum)
for num in range(0,len(vec_lsi)):
slist[vec_lsi[num][0]]=vec_lsi[num][1]
wlist=wlist+slist
writer.writerow(wlist)
k=k+1
if(k%1000==0):
print k,time.ctime()
ifile.close()
wfile.close()
print "fin",time.ctime()
def main(model_,bnum_,tnum_,train_,pas_): from gensim import corpora, models, similarities import csv import textedit import time model=str(model_) bnum=int(bnum_) tnum=int(tnum_) train=str(train_) pas=str(pas_) print "mkcorpus_start",time.ctime() #remove subrev subfile=open(pas+"subrev_1000.csv","r") subdata=csv.reader(subfile) sublist=[] for line in subdata: stoplist.append(line[0]) subfile.close() ##remove stoplist stopfile=open(pas+"stopwords/over4word.csv","r") stopdata=csv.reader(stopfile) stoplist=[] for line in stopdata: stoplist.append(line[0]) stopfile.close() stopset=set(stoplist) ##make documents dfile=open(pas+train+".csv","r") ddata=csv.reader(dfile) ddata.next() dnum=0 dlist=[] for line in ddata: if(line[0] not in sublist): te=line[5] doc=textedit.textedit(te) dlist.append(doc) dnum=dnum+1 if(dnum%10000==0): print dnum, dfile.close() print "dfile fin",time.ctime() texts = [[word for word in document.lower().split()] for document in dlist] print "text fin",time.ctime() texts = [[word for word in text if word not in stopset] for text in texts] dictionary = corpora.Dictionary(texts) print "dictionary fin",time.ctime(),len(dictionary.token2id) dictionary.filter_extremes(no_below=bnum) print "dictionary cut fin",time.ctime(),len(dictionary.token2id) dictionary.save(pas+train+"_o4b"+str(bnum)+"t"+str(tnum)+".dict") dictionary.save_as_text(pas+train+"_o4b"+str(bnum)+"t"+str(tnum)+"_text.dict") corpus=[dictionary.doc2bow(text) for text in texts] corpora.MmCorpus.serialize(pas+train+"_o4b"+str(bnum)+"t"+str(tnum)+".mm", corpus) print "mk_corpus fin",time.ctime()
header.append("t"+str(num).zfill(int(topic_num)/10))
wfile=open(pas+"nNVrev_LSI_t"+str(topic_num)+".csv","wb")
writer=csv.writer(wfile)
writer.writerow(header)
"NVreview.csv:[review_id,user_id,business_id,stars,date,texts]"
ifile=open("D:/Lresult/NVreview.csv","r")
idata=csv.reader(ifile)
idata.next()
k=0
for line in idata:
wlist=[]
wlist.append(line[0])
wlist.append(line[2])
doc=textedit.textedit(line[5])
vec_bow = dictionary.doc2bow(doc.lower().split())
vec_lsi = lsi[vec_bow]
###
slist=[0]*int(topic_num)
for num in range(0,len(vec_lsi)):
slist[vec_lsi[num][0]]=vec_lsi[num][1]
wlist=wlist+slist
writer.writerow(wlist)
k=k+1
if(k%1000==0):
print k,time.ctime()
ifile.close()
wfile.close()
print "fin",time.ctime()