Internati
o
nal
Journal of Ele
c
trical
and Computer
Engineering
(IJE
CE)
V
o
l.
5, N
o
. 5
,
O
c
tob
e
r
201
5, p
p
. 1
180
~118
7
I
S
SN
: 208
8-8
7
0
8
1
180
Jo
urn
a
l
h
o
me
pa
ge
: h
ttp
://iaesjo
u
r
na
l.com/
o
n
lin
e/ind
e
x.ph
p
/
IJECE
Online Crowds Opinion-Mining it
to An
alyze Current Tren
d: A
Review
Haritha Akki
neni*,
P.
V.
S.
Lakshmi
*
*, B
.
Vi
jay B
a
bu
***
* Department of
Computer Scien
ce
a
nd
Engineering, KL University
, India
** Dept of
Infor
m
ation Technolo
g
y
, PVP Siddhar
t
ha
Institute of
Technolog
y
,
Vijay
a
w
a
da, India
*** Dept of
CS
E, KL University
, India
Article Info
A
B
STRAC
T
Article histo
r
y:
Received Apr 30, 2015
Rev
i
sed
Jun
13,
201
5
Accepted
Jun 30, 2015
Online presen
ce of the user has incr
eased
, there is a huge gr
owth in the
number of activ
e users and thus the vol
ume of
data created on
the online
social n
e
tworks
is massive. Much ar
e
concentrating on th
e Inter
n
et Lingo.
Notably
most of
the data on
the s
o
cial
networking
sites is made pu
blic which
opens doors for
companies, r
e
searchers
a
nd
analy
s
t
to co
llect an
d analy
z
e th
e
data. We have h
uge volume of opinione
d data
av
ailable on th
e web we hav
e
to m
i
ne it so that we could get s
o
m
e
in
teresting
results out of it with could
enhance th
e decision making process. In
order
to
anal
yz
e th
e cur
r
e
nt s
cen
ario
of what peop
le
are th
inking fo
cus is
shifted to
wards opinion
mining. This
s
t
ud
y
pres
ents
a
s
y
s
t
em
atic l
i
t
e
r
a
ture rev
i
ew tha
t
contains
a co
m
p
rehens
ive
overview of co
mponents of opinion mining,
subjectivity
of
data, sources of
opinion, th
e process and how do
es it le
t one an
al
yz
e the curr
ent t
e
ndenc
y o
f
the online crowd
in a particul
ar c
ontext
.
Different perspectiv
es from different
authors reg
a
rding the
above
scenario
have
been pr
esented
.
Resear
ch
chal
lenges
and
differen
t
app
lic
a
tions tha
t
wer
e
develop
e
d with
the m
o
tive
opinion
mining
are also
d
i
scussed.
Keyword:
Cr
ow
d So
ur
cing
Micr
o
Blogg
ing
Opi
n
i
o
n M
i
ni
n
g
Social Net
w
orks
Wo
r
d
Se
nse
Di
sam
b
i
guat
i
o
n
Copyright ©
201
5 Institut
e
o
f
Ad
vanced
Engin
eer
ing and S
c
i
e
nce.
All rights re
se
rve
d
.
Co
rresp
ond
i
ng
Autho
r
:
Hari
t
h
a
A
k
ki
n
e
ni
,
R
e
search
Sc
ho
l
a
r, Dept
of
C
S
E, KLU
,
7
4
-1
3
/
1
-
2
4
A
,
Plo
t
no
:
2
4
6
,
New
RTC C
o
lony,
V
i
j
a
yaw
a
d
a
, An
dhr
a Pr
ad
esh, In
d
i
a
52
000
7.
Em
a
il:ak
k
i
n
e
n
i
h
@
g
m
ail.co
m
1.
INTRODUCTION
The i
n
crea
sed
penet
r
at
i
on
of
t
h
e i
n
t
e
rnet
am
on
g t
h
e f
o
l
k
s
has m
a
de it
a
wi
de i
n
di
spe
n
s
i
bl
e chan
nel
fo
r pe
o
p
l
e
t
o
com
m
uni
cat
e.
There i
s
m
u
ch
gr
owt
h
o
f
t
h
e
W
o
rl
d
W
i
de
Web
n
o
t
onl
y
i
n
i
t
s
si
ze but
al
so i
n
term
s
o
f
serv
ices an
d
th
e co
n
t
en
t th
at is b
e
ing
prov
id
ed
. T
h
ere is a huge growth in
the
num
b
er of active users
an
d thu
s
t
h
e
vo
lu
m
e
o
f
d
a
ta created on
t
h
e on
lin
e so
cial n
e
twork
s
is
massiv
e
. "Twitter, a
p
opu
lar micro
-
bl
o
ggi
ng
ser
v
i
ce, has
2
88 m
i
l
l
i
on m
ont
hl
y
act
i
v
e user
s,
w
ho
p
o
st
ab
o
u
t
50
0 m
i
ll
i
on t
w
eet
s a day
"
[1]
.
Thi
s
i
t
s
el
f depi
ct
s t
h
at
t
h
e en
o
r
m
ous am
ount
of
us
er
gene
rat
e
d
co
nt
ent
i
s
bei
n
g
g
e
nerat
e
d e
v
ery
day
.
There
has
bee
n
a dri
f
t
i
n
h
o
w t
h
e i
n
f
o
r
m
at
i
on i
s
bei
ng m
a
nage
d and s
h
ar
ed
. Fr
om
onl
y
just
con
s
um
i
ng t
h
e
avai
l
a
bl
e co
nt
e
n
t
t
o
a
n
not
at
i
n
g i
t
an
d
gene
ra
t
i
ng
new
i
n
f
o
r
m
at
i
on. T
h
ere
can
be
di
ffe
ren
t
way
s
lik
e co
mmen
t
s o
n
th
e ex
itin
g
inform
atio
n
,
boo
k
m
ark
p
a
g
e
s, prov
ide ratin
g
s
, sh
are th
eir id
eas with
comm
unity at large
.
O
n
lin
e pr
e
s
en
ce
o
f
th
e
u
s
er
ha
s
increas
ed,
Much a
r
e c
o
nc
entrating
on t
h
e Int
e
rnet
Li
ng
o.
N
o
t
a
bl
y
m
o
st
of t
h
e
dat
a
o
n
t
h
e
soci
al
net
w
or
ki
ng
si
t
e
s i
s
m
a
d
e
p
ublic w
h
ich op
ens do
or
s
fo
r
com
p
an
ies, r
e
sear
ch
er
s
and a
n
alyst to collect and a
n
alyze the data. Resear
che
r
s
are m
onitoring on the trending topics, m
e
mes and
so
m
e
n
o
t
ab
le ev
en
ts i
n
clud
ing
po
litical ev
en
ts [2
], sto
c
k
mark
et flu
c
t
u
atio
n
s
[3
],
d
i
sease ep
id
em
ics [4] etc to
source out
the voice of
crowd
and observ
e the present te
nde
n
cy of
online c
r
owds toward
s
a p
a
rticu
l
ar issu
e.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
I
J
ECE Vo
l. 5
,
N
o
. 5
,
O
c
tob
e
r
20
15
:
118
0
–
11
87
1
181
We
have
h
u
g
e
vol
um
e of
opi
ni
o
n
ed
dat
a
a
v
ai
l
a
bl
e o
n
t
h
e
we
b
we
have
t
o
m
i
ne i
t
so t
h
at
we
co
ul
d
get
s
o
m
e
i
n
t
e
rest
i
ng
res
u
l
t
s
o
u
t
of
i
t
wi
t
h
co
u
l
d en
ha
nce t
h
e
deci
si
o
n
m
a
ki
ng
pr
ocess
.
Reco
gn
izing
an
d ev
al
u
a
tion o
f
v
a
riou
s
v
i
ewpo
in
ts h
a
s
g
o
t
its
p
r
acticality in
v
a
rious do
m
a
in
s.
Decision
m
a
k
e
rs in gov
ern
m
en
ts an
d po
litical en
tities n
e
ed
to kno
w how th
e pub
lic reto
rt to th
e
d
e
cisio
n
s
t
a
ken by
t
h
em
, fun
d
i
n
g age
n
ci
es need t
o
g
a
uge t
h
ei
r suc
cess, b
u
si
ness
need t
o
kn
o
w
what
pe
o
p
l
e
t
h
i
nk o
f
th
eir p
r
od
u
c
ts, Th
e i
m
p
act o
f
a research
ers work
can
b
e
calcu
lated
b
y
h
i
ring
co
mmitte
es o
f
un
iv
ersities and
research
in
stitutio
n
s
. Su
ch
postin
g
s
h
a
v
e
also
m
o
b
ilized
masses for po
litical ch
an
g
e
s such
as th
o
s
e
h
a
p
p
e
n
e
d
in
so
m
e
Arab
co
un
tries in
2
0
1
1
. It h
a
s thu
s
b
eco
m
e
a n
ecessity to
co
llect
an
d
stud
y op
i
n
ion
s
on
th
e
Web.In
or
der
t
o
a
n
al
y
z
e t
h
e c
u
r
r
ent
sc
enari
o
of
w
h
at
peo
p
l
e
a
r
e t
h
i
n
ki
n
g
foc
u
s i
s
s
h
i
f
t
e
d
t
o
war
d
s
opi
ni
o
n
m
i
ni
ng
.
Th
is st
u
d
y
p
r
esen
ts a
systematic li
teratu
re re
vi
ew t
h
at
co
nt
ai
ns a
c
o
m
p
rehe
nsi
v
e
ove
r
v
i
e
w
o
f
com
pone
nt
s
of
o
p
i
n
i
o
n m
i
ni
n
g
,
su
b
j
ect
i
v
i
t
y
of
dat
a
,
h
o
w
i
t
rel
a
t
e
s t
o
sent
im
ent
anal
y
s
i
s
, so
urce
s
of
o
p
i
n
i
o
n
.
Th
e pro
cess and
how
do
es it let o
n
e
an
alyze the curre
n
t tendency
of t
h
e
o
n
l
i
n
e cr
ow
d i
n
a part
i
c
ul
ar c
o
nt
ext
.
Diffe
re
nt per
s
pectives f
r
o
m
diffe
re
nt
authors re
garding t
h
e above scenar
io ha
s bee
n
prese
n
ted. Res
earch
chal
l
e
ng
es a
n
d
di
f
f
ere
n
t
a
ppl
i
cat
i
ons t
h
at
we
re
devel
ope
d
w
i
t
h
t
h
i
s
m
o
t
i
v
e opi
ni
o
n
m
i
ni
ng
are al
s
o
di
scus
sed.
2.
LITERATU
R
E
REVIE
W
2.
1 O
p
i
n
i
o
n
Mi
ni
ng
It
em
erges fr
o
m
t
h
e basi
c
fi
el
d o
f
t
e
xt
m
i
ni
ng
[
5
]
,
a s
u
bse
t
gene
ral
l
y
t
e
xt
i
n
unst
r
uct
u
re
d f
o
rm
at
t
o
or
ga
ni
ze i
t
i
n
a pr
ope
r w
a
y
an
d t
o
m
i
ne som
e
usef
ul
i
n
f
o
r
m
at
i
on fr
om
it
. Opi
n
i
o
ns are t
h
e key
i
n
fl
ue
n
cers o
f
ou
r be
havi
or w
h
i
c
h su
p
p
l
e
m
e
nt
s
t
h
e
deci
si
o
n
m
a
ki
ng p
r
oc
ess
[
6
]
.
Op
i
n
ion
:
An
op
in
ion
is a
q
u
i
n
t
up
le,
(ei; aij
;
oo
ijk
l; hk
; tl),
wh
ere ei is t
h
e
n
a
m
e
o
f
an
en
tity,
aij is a
n
as
pect
of ei,
o
o
i
j
k
l is th
e
orien
t
atio
n
of th
e
o
p
i
n
i
on
ab
ou
t
asp
ect aij ofen
tity ei,
h
k
is th
e op
in
i
o
n ho
ld
er,
an
d tl is th
e ti
me wh
en
th
e op
in
ion
is exp
r
essed
b
y
h
k
.
The
opi
ni
o
n
o
r
i
e
nt
at
i
on
o
o
i
j
k
l
can
be p
o
si
t
i
v
e, ne
gat
i
v
e
or
neut
ral
,
o
r
be ex
p
r
esse
d wi
t
h
di
ffe
rent
streng
th
/in
ten
s
ity lev
e
ls.
To ca
rry
o
u
t
t
h
e ge
neral
pr
oce
ss o
n
e
nee
d
s t
o
pe
rf
orm
t
h
e f
o
l
l
o
wi
n
g
t
a
s
k
s:
Fi
gu
re
1. Ta
sks
i
n
v
o
l
v
e
d
2.
2 Com
p
o
n
e
n
ts of
Opi
n
i
o
n
Mi
ni
n
g
Sou
r
ce of th
e op
in
ion
:
Obt
a
i
n
i
ng
p
u
b
l
i
c
and c
ons
u
m
er opi
ni
o
n
s
has l
o
n
g
bee
n
a hu
ge t
r
a
d
e
i
t
s
el
f for m
a
rk
et
i
ng,
pu
bl
i
c
relatio
n
s
, and
p
o
litical ca
m
p
aig
n
co
m
p
an
ies. Th
e scen
ario
of ask
i
ng
o
t
h
e
rs
fo
r th
ei
r
p
e
rsp
ectiv
e
h
a
d
sh
i
f
ted
t
o
wa
rds
o
n
l
i
n
e
cro
w
d s
o
urci
ng
[7]
.
T
h
ere
are di
f
f
ere
n
t
f
o
rm
s l
i
k
e con
duct
i
n
g s
u
r
v
ey
s, o
p
i
n
i
o
n
pol
l
s
and
f
o
cu
s gr
oup
s.
User
Ge
ne
rat
e
d C
ont
e
n
t
:
B
l
o
ggi
ng
, M
i
cr
o
b
l
o
g
g
i
n
g
an
d
so
ci
al
net
w
or
ks
are t
h
e
m
o
st
p
o
p
u
l
a
r
f
o
rm
s
o
f
UG
C [8
].
Entity
Ex
trac
tion
and
G
r
ouping
Aspect
Ex
trac
tion
and
g
r
ouping
Opinion
Ho
ld
e
r
and
time
ex
trac
tion
Opinion
Q
u
intuple
Gen
e
rat
i
on
Aspect
Sentiment
Classif
i
cat
i
on
Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE
ISS
N
:
2088-8708
Onl
i
n
e
C
r
ow
ds
O
p
i
n
i
o
n
-
Mi
ni
ng
i
t
t
o
A
n
al
yz
e C
u
rre
nt
Tre
n
d:
A Revi
ew
(
H
ari
t
h
a Akki
nen
i
)
1
182
Questi
on
-A
ns
wer
Data
bases (eg
.
Yah
o
o
An
swers
,
As
k.c
o
m
)
Dig
ital v
i
deo (You
tub
e
, Vimeo
)
B
l
ogs (B
l
o
gge
r
,
W
e
e
b
l
y
)
Micro
B
log
s
(Tu
m
b
l
r,
Twitter)
Po
dcast
i
n
g
(i
T
une
s)
ReviewSites
(Yelp,
Tri
p
Advisor)
Social Net
w
orking
(Face
book, MySpace)
W
i
k
i
s
(W
ik
ep
ed
ia
)
Al
o
ng
wi
t
h
t
h
e di
gi
t
a
l
cont
e
n
t
t
h
e U
G
C
c
a
n be a c
o
m
b
inat
i
on
of
o
p
en
sou
r
ce,
free s
o
ft
ware a
n
d
flex
ib
le licensin
g
or
related
ag
reem
en
ts to
fu
rt
h
e
r
re
d
u
c
e th
e
b
a
rricad
es to
th
e co
llectiv
e wo
rk
. All these can
be
t
h
e
s
o
u
r
ce whe
r
e we
c
o
ul
d gat
h
e
r
t
h
e
pe
rspect
i
v
e
s
of v
a
ri
o
u
s peo
p
l
e
.
2.
3 F
a
ct
a
nd
Opi
n
i
o
n
Subjec
tivi
ty analysis
:
The subjective
sentence expresses
som
e
percept
i
o
ns,
bel
i
e
fs o
r
pe
rso
n
al
feel
i
ngs
. The
r
e are
m
a
ny
fo
rm
s of S
u
bje
c
t
i
v
e ex
pres
si
o
n
s l
i
k
e
al
l
e
gat
i
ons
,
desi
res
,
be
l
i
e
fs, s
u
spi
c
i
o
n
s
, a
n
d
spec
ul
at
i
ons
Objec
t
i
v
e an
a
l
ysi
s
:
The objective
sentence c
o
nce
n
trates
on the
facts.T
h
e exist
i
ng re
searc
h
c
once
n
trates
on the factual
in
fo
rm
atio
n
wh
ich is
wid
e
ly
u
s
ed
in th
e area of
Info
rm
atio
n
Retriev
a
l.
No
w peop
le are
m
o
v
i
n
g
t
o
wards wh
at
peo
p
l
e
feel
an
d opi
ni
o
n
ho
o
d
det
e
rm
i
n
at
i
on. O
p
i
n
i
o
n h
o
o
d
det
e
rm
i
n
at
ion
i
s
di
vi
de
d i
n
t
o
t
w
o
s
u
b
t
a
sks
Sub
j
ectiv
ity cl
assif
i
catio
n
and
o
p
i
n
i
on
po
lar
ity classif
i
cati
o
n wh
ich
is i
n
tu
rn
r
i
pp
ed
i
n
to
o
p
i
n
i
on
an
d non
opi
nion, positive and ne
gative respectiv
ely.
[9] The m
e
thod they propos
e
is
using m
i
nim
u
m
cuts to produce
su
bj
ectiv
e ex
tracts fro
m
th
e t
e
x
t
. Th
e
work
h
a
s b
e
en
fo
cused
in
th
e sen
t
en
ce lev
e
l sub
j
ectiv
ity ex
tracti
o
n. Th
e
pr
ocessi
ng
o
v
e
r
hea
d
of
l
a
r
g
e am
ount
o
f
dat
a
can be
el
i
m
i
n
at
ed
[
1
0]
.
Fi
gu
re
2.
O
p
i
n
i
o
n
h
o
o
d
det
e
r
m
i
n
at
i
on
2.
4 Appr
o
a
ch
es
Ad
ap
ted
Th
e
fo
llowing
are th
e app
r
o
a
ch
es ad
ap
ted
for op
in
ion
m
i
n
i
ng
: [11
]
1
.
Heu
r
istics:
Th
is ap
pro
ach
can
p
r
od
u
c
e the resu
lts with
i
n
a realistic ti
m
e
fram
e
. Th
ey are lik
ely to
p
r
o
d
u
ce th
e
resu
lts t
h
em
sel
v
es
bu
t are m
o
stly u
s
ed
with
t
h
e
o
p
tim
ized
alg
o
rith
m
s
.
2. Di
sc
ou
rse St
ruct
ure:
Foc
u
ses
on t
h
e gi
ve
n t
e
xt
t
h
at
just
com
m
uni
cat
es a
m
e
ssage, a
nd l
i
n
ki
ng i
t
t
o
ho
w t
h
at
m
e
ssage
co
nstru
c
ts a social reality o
r
view of t
h
e
world
.
3. C
o
arse-grai
n
ed Analysis:
The tasks s
u
c
h
as retrieving th
e subjective doc
um
ents from a collect
ion, m
o
st
recent
work on this
to
p
i
c h
a
s fo
cu
sed
o
n
classificatio
n
o
f
en
tire d
o
c
u
m
en
t
b
y
ov
erall p
o
s
itiv
e an
d
n
e
g
a
tiv
e
po
larity.
4. Fi
ne g
r
ai
ne
d
A
n
al
y
s
i
s
:
Task
s su
ch
as
d
e
term
in
in
g
the attitu
d
e
of a
p
a
rticu
l
ar
p
e
rso
n
on
a
p
a
rticular to
p
i
c. It is
a term
lev
e
l
an
alysis as it iden
tifies wh
et
h
e
r th
e term
is p
o
sitiv
e o
r
n
e
g
a
ti
v
e
o
r
ien
t
ed
.
5.
As
pect
l
e
vel
ap
pr
oac
h
:
Opi
n
i
o
ns c
o
nsi
s
t
e
d
of t
a
rget
s
and
the a
s
pects
associated wit
h
them
.
Opinionhood
de
te
rmination
Subjectivity
Classific
a
tion
Opinion
Non
‐
opinion
Opinion
Polarity
Classific
a
tion
Pos
i
tive
Negative
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
I
JECE Vo
l. 5
,
N
o
. 5
,
O
c
tob
e
r
20
15
:
118
0
–
11
87
1
183
6. Key
wo
r
d
a
n
aly
s
is:
This a
p
proa
ch classifies text
by affect cat
egor
i
e
s
based
on
t
h
e
pre
s
en
ce of
u
n
am
bi
guo
us a
ffect
w
o
r
d
s su
ch as
h
a
pp
y,
sad, afraid
, an
d bor
ed.
7
.
Con
c
e
p
t an
alys
is
:
It
conce
n
t
r
at
es
on sem
a
nt
i
c
anal
y
s
i
s
of t
e
xt
thr
o
ug
h t
h
e use
of we
b o
n
t
o
l
o
gi
es or sem
a
nt
i
c
net
w
o
r
ks.
The c
o
ncept
u
al and a
ffective
inform
ation ass
o
ciated
with
n
a
tu
ral
langu
ag
e o
p
i
n
i
on
s
are
agg
r
eg
ated.
3.
R
E
SEARC
H M
ETHOD
Phase
1:
Opini
o
n ext
r
action.
Phase
2:
Pr
oce
ssi
ng
usi
n
g
di
f
f
ere
n
t
t
ech
ni
q
u
e
s
Phase
3:
Anal
yzing them
to
assess th
e curren
t tren
d
Fi
gu
re
3.
Sc
he
m
a
t
i
c
proce
d
u
r
e o
f
Opi
n
i
o
n
M
i
ni
ng
3.1 The
list
of
review ar
ticles sur
v
eye
d
Thi
s
t
a
bl
e
co
nt
ai
ns
vari
o
u
s
fi
ndi
ng
s
on
o
p
i
n
i
on m
i
ni
ng
fr
o
m
di
fferent
a
u
t
h
o
r
s
pe
rspect
i
v
es. It
cl
ea
rl
y
states the m
e
thodology, al
gori
thm
used,
data
sources
an
d t
h
e res
u
l
t
s
an
d c
oncl
u
si
o
n
s t
h
at
we
re
dra
w
n
.
Tabl
e 1. Art
i
c
l
e
s
R
e
vi
ewe
d
Ref.
No
Study
Year
Data
scope
Algorit
h
mn
Met
h
odology used
Dat
a
Source
Conclusion Result
[12]
Yao
W
u
et al
2015
Hotel
r
e
view data
and
restau
ran
t
s
Machine
L
ear
ning
Co
m
b
ination of
aspect-based
opinio
n
m
i
ning
an
d
collaborative
filtering to
collectively le
arn
users’ pref
erences
on dif
f
er
ent
aspects.. A
unif
i
ed
probabilistic m
ode
l
Factorized L
a
tent
Aspect Model
(FLA
ME
) h
a
s b
e
e
n
pr
oposed.
Trip
Advisor
hotel
review
data6 and
Yelp
rev
i
ew
data7
Aspect-based
opinio
n
m
i
ning
techniques,
aspect-based
review
su
mm
a
r
ization for
pr
oducts pr
oduct
level
su
mm
a
r
ization-
r
e
view-
level
analysis. P
r
edicts
the pref
erences
based on pr
evious
pref
erences
FLAM
E
outper
f
orm
s
other
baseline m
odels
on
all the
m
easures
except RMSE
. The
gain is especially
significant
co
m
p
a
r
ed to
LRR+PMF
where
ther
e are about 90%
im
p
r
ovem
e
nt on _I
and 40%
im
p
r
ovem
e
nt.
[13]
Z
h
eng
Xiang et
al
2015
Hotel
guest
experience
and
satisfaction
Machine
L
ear
ning /
statistical
analysis
Classify
ing large
am
ount of online
custo
m
e
r
r
e
views,
assess the quality
of
these data, as
well
as identify inherent
r
e
lationships
between two
do
m
a
ins of
var
i
ables in hotel
m
a
nagem
e
nt
thr
ough text
analytics.
consu
m
er
reviews
extracted
fr
o
m
E
xpedia.
co
m
Is study
sets an
exa
m
pl
e f
o
r the
develop
m
ent of
business analytics
in hospitality
m
a
r
k
eting and
m
a
nagem
e
nt Or,
at least,
these
findi
ngs can be
seen as testable
propositio
ns
der
i
ved fr
o
m
the
analysis.
T
h
e under
l
y
i
ng
f
actors representing
a set of only
34
words can explain
near
ly
63% of the
total variance in
guest satisfaction,
which considerably
exceeded the
acceptable range.
User
Gen
e
rat
e
d
Co
n
t
e
n
t
•
F
ou
rms,Wikis,
Blogs,Social
Netw
ork
i
ng
sit
e
s
Pre
‐
Processin
g
•
Toenize,
Re
mo
ve
st
op
wo
r
d
s,
Stemming
Tex
t
Minin
g
&
In
f
o
rmat
ion
Processin
g
•
A
ssociat
ion
ules
,
C
l
us
tering
,
•
C
lassif
i
cat
i
on
,
C
a
teg
o
rization
In
f
o
rmat
ion
Visualization
•A
n
a
l
y
s
i
s
to
assess
on
current
trend
•S
t
r
a
t
e
r
g
i
c
dec
i
s
i
on
mak
i
ng
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
ECE
I
S
SN
:
208
8-8
7
0
8
Onl
i
n
e
C
r
ow
ds
O
p
i
n
i
o
n
-
Mi
ni
ng
i
t
t
o
A
n
al
yz
e C
u
rre
nt
Tre
n
d:
A Revi
ew
(
H
ari
t
h
a Akki
nen
i
)
1
184
[14]
W
e
nting
Tu
,et al
2015
Shanghai
ma
r
k
e
t
I
ndex (
S
HI)
Rule
based
m
e
thods
Ref
e
rence ti
m
e
(RT) inform
ation in
the form
ation of
future predictions
constr
ucted a
pr
ediction m
odel
that uses the RT
inform
a
tion.
50,
169
m
i
cr
oblogs
of investor
s
posted on
Sina
W
e
ibo.
Pr
edictions m
a
de
using r
e
fer
e
nce
ti
m
e
(RT)
Ti
m
e
-sensitive
m
odel,
which is
based on RT
tags
per
f
orm
s
better
than
Baseline and
Traditional m
odels
.
[15]
Subasish
Das et al
2015
Capital
Bike share
of
W
a
shington
DC
L
e
xicon
Apply
i
ng sentim
ent
analysis
Twitter
To get subjective
inform
ation T
e
xt
categorization
with valence
annotation was
used.
The positive
r
e
sponses towar
d
s
the current s
y
ste
m
wer
e
higher
in
fr
equency
than the
negative ones
[16]
T
huy
T.B.
L
uong et
al.
2015
Rail
tr
ansit
line fro
m
travelers
perspective
Opinion
L
e
xicon
in English
pr
ovided
by
Hu and
Liu
E
xplor
ator
y
analysis and word
clustering analysis
was per
f
orm
e
d
Tweets
f
r
o
m
Ma
y
1st 2014 to
October
1st
2014
co
mm
ute
r
Interactive online
interf
ace which
display
s
and
m
onitor
real
-ti
m
e
f
eedback and
senti
m
ent helpful
for transit service
pr
ovider
s
.
M
onday
s
had
m
o
re
positive tweets40%
Tu
esd
a
y m
o
re
negative
Tweets-55%
[17]
Yan
Zhao et al
2015
Chinese
hotel
rev
i
ews.
Machine
lear
ning
to extract
aspects.
D
i
ffe
r
e
nt
dim
e
nsions of
f
eatures, f
eature
r
e
pr
esentation
m
e
thods and
classif
i
ers to
analyze the
integra
l
ef
f
ect of
aspect
extraction.
Online
rev
i
ews
ME is
th
e b
e
st
m
a
chine lear
ning
m
e
thod for
aspect
extr
action of
Chinese hotel
rev
i
ews
T
h
e highest
accurac
y
of
differ
e
nt super
v
ise
d
lear
ning m
e
thod
are
93.
47% (
M
E
)
,
87.
82% (
N
B)
and
86.
34% (
S
VM
).
[18]
Chetashr
i
hadane et
al
2015
M
obile
do
m
a
in
Machine
lear
ning
(SVM
)
co
m
b
ined
with
do
m
a
in
specif
i
c
lexicons.
T
w
o-
step m
e
thod
aspect classif
i
cation
followed by
polar
ity
classif
i
cation
Online
rev
i
ews
A set of
techniques like
linear kernel f
o
r
aspect
classif
i
cation and
polar
ity
identification of
pr
oducts
Linear Kernel
backed up with
ma
x
i
mu
m a
c
c
u
r
a
c
y
.
The accura
cy
achieved was
78.
05% testing was
done usin
g 41
rev
i
ews.
[19]
Rahul
Tejwani
et al
2014
Academ
ic
L
e
xicon
based
f
eatures
T
h
ay
es M
odel
of
hu
m
a
n e
m
otion
Yelp
Review
dataset
Classify text in
two dim
e
nsional
space on polarity
and intensity
Polarity:
T
m
ean
accurac
y
81.60%
+/- 1.92% intensity:
m
e
an ac
curac
y
67.
14% +/-
1.
22%.
[20]
Nathan
Aston et
al
2014
Sander
s
Cor
pus
STS-Gold
and Senti
Str
e
ngth
Machine
L
ear
ning
Modif
i
ed Balanced
W
i
nnow to tr
ain o
n
pr
e-
labeled
instances using a
str
e
am
ing algor
ith
m
and pr
ocessing
the
m
in real ti
m
e
.
public
datasets
Twitter,M
y
Space,
Youtube,
B
BC, Digg
Rem
oved all but
the top f
eatures
when per
f
orm
i
ng
classif
i
cation.
Achieved the
highest accuracy o
f
87.
5% with
STS_Gold on 5
gr
am
s
r
e
pr
esentation.
3.
2 Appl
i
c
a
t
i
o
ns
De
vel
o
ped
1. I
F
eel:
It is a
web
applicatio
n
th
at t
h
at
det
ect
s se
nt
i
m
ent
s
i
n
a
n
y
f
o
rm
of
t
e
xt
i
n
c
l
udi
n
g
u
n
st
r
u
ct
ure
d
s
o
ci
al
m
e
di
a dat
a
. A
com
b
i
n
ed M
e
t
h
o
d
was
de
ve
l
ope
d.
Usi
n
g t
h
i
s
t
ool
t
h
e
us
er can a
n
al
y
s
e t
h
e gi
ve
n dat
a
wi
t
h
com
b
ined m
e
thod as
well as
the se
v
e
n ex
istin
g
sen
tim
en
t
an
alysis m
e
th
o
d
s l
i
k
e:
Se
nt
i
W
o
r
dNet
,
Em
ot
i
c
on
s,
PANAS-t,
SASA,
Ha
ppiness
Inde
x,
Se
n
tic-Net, an
d Sen
tiStreng
t
h. [21
]
2. Soci
al
Me
ntion:
It is a s
o
cial media search and a
n
alysis plat
form
that aggregates
user
ge
nerat
e
d c
ont
e
n
t
fr
om
across
t
h
e u
n
i
v
e
r
se i
n
t
o
a si
ngl
e st
re
am
of i
n
f
o
rm
ati
on.
It
hel
p
s
u
s
i
n
t
r
ac
ki
n
g
a
n
d
m
easuri
ng
w
h
at
peo
p
l
e
say
.
[
22]
3. Sentime
n
t140
Sentim
ent 140 is used to disc
ove
r the senti
m
ent of
a brand, product, or t
opi
c on Twitter. Classifiers
are
build usi
n
g m
achine learning al
gorithm
s
. [23]
4.
Opini
o
n Finder
:
It identifies s
u
bjective se
nt
ences and m
a
rks
vari
ous as
pects of subje
c
tiv
ity in these sentences
,
in
clu
d
i
n
g
th
e so
urce
o
f
t
h
e sub
j
ectiv
ity and
words th
at are
in
clu
d
e
d
i
n
phrases exp
r
essing
po
sitiv
e
o
r
neg
a
tiv
e
sen
tim
en
ts. [24]
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
I
J
ECE Vo
l. 5
,
N
o
. 5
,
O
c
tob
e
r
20
15
:
118
0
–
11
87
1
185
5. T
r
ack
u
r
It is a social
media
m
onitori
ng se
rvice whic
h ena
b
les us to gene
rate speci
fic searches t
o
run across a
n
u
m
b
e
r of so
cial
m
e
d
i
a p
l
atfo
rm
s. Track
s
po
sts fro
m
Twitter, Faceb
ook
,
Red
d
it, Delici
o
u
s
, Goo
g
l
e+,b
log
s
,
vi
de
os i
n
cl
u
d
i
n
g
Yo
uT
u
b
e, a
n
d
ph
ot
o
s
i
n
cl
u
d
i
n
g Fl
i
c
k
r
[2
5
]
.
Tabl
e
2. T
o
ol
s
use
d
f
o
r a
part
i
c
ul
ar m
e
t
hod
Met
h
od Tools
Used
Sear
ch
Google,
T
opsy
,
So
cial M
e
ntion,
T
w
itter
,
FB
M
onitor
i
ng
Google Aler
t,
T
r
ackur
,
Atte
ntio,
E
ngagor
,
Vir
a
l Heat,
Radian6,
Hub Spot
T
r
ack Per
f
orm
a
nce
Google Analy
tics,
Post
Rank,
Str
u
tta,Your
Pr
ofile,
Swix
,
Klout,
Peerindex, Gr
ader
Searc
h
sol
u
t
i
o
ns are
p
r
o
v
i
d
e
d
by
Go
o
g
l
e
, f
o
r
fol
l
o
wi
n
g
n
e
ws,
bl
o
g
post
s
, vi
deos a
n
d i
m
ages t
h
e
Social Mention,
Addict-o-matic etc
p
l
ay a
p
r
o
m
in
en
t ro
le.Techno
rati.com
facil
itates in
search
i
n
g b
l
og
s, and
B
o
t
B
ox
, fo
r on
l
i
n
e
ne
ws
c
o
nt
ent
,
i
n
cl
udi
n
g
bl
o
g
s .
An
alytical to
o
l
s fo
r
Twitter
are Tweet Arch
iv
ist,
14
0k
it
an
d TweetDeck
, is
an agg
r
eg
atio
n too
l
.
Twap
p
e
rk
eep
e
r, is an
API
fo
r track
i
n
g Twitt
er activ
ities.
An
o
p
en
-s
ou
rc
e com
m
a
nd-l
i
n
e t
ool
t
o
f
u
rt
h
e
r p
r
oce
ss d
a
t
a
-Gawk
e.g. tweet statistics
an
d
m
e
trics,
and t
h
e
ope
n s
o
u
r
ce
vi
sual
i
z
a
t
i
on t
o
ol
s
Wo
r
d
l
e
, u
s
ed al
s
o
f
o
r
Y
ouT
u
b
e
vi
deo t
a
gs, a
n
d
Gep
h
i
f
o
r
vi
s
u
al
i
z
i
ng
n
e
two
r
k
s
,
u
s
ed fo
r e.g
.
Twitter d
a
tasets and
b
l
og
. Th
e exp
a
n
s
ion
of to
o
l
s
co
n
c
ern
i
ng
so
cial
med
i
a ex
amin
in
g
is swift and
th
ere is a am
p
l
e sco
p
e
fo
r
m
a
ny
new
o
ffs
p
r
i
n
gs
t
o
a
r
i
s
e.
[2
6]
3.
3 Rese
arch Ch
al
l
e
nges
Core
ference
r
e
soluti
on:
Th
e
p
r
o
cess
o
f
find
ing
all exp
r
essi
on
s th
at refer to th
e same en
tity in
a tex
t
It is im
p
o
r
tan
t
ch
allenge
for op
in
i
o
n
m
i
n
i
ng
as w
ithout co
n
s
id
ering
a g
r
eat d
eal
o
f
op
in
ion
inform
a
tio
n
w
ill b
e
lo
st, an
d
o
p
i
n
i
on
s
m
a
y
b
e
assi
g
n
e
d
t
o
wro
n
g
en
tities. [2
6
]
N
e
gat
io
n ha
nd
lin
g
:
Th
e
g
r
amm
a
ti
cal ru
les
u
s
ed
in
th
e tex
t
u
a
l
d
a
ta often con
t
ain
n
e
g
a
tion
s
u
s
ed
in tex
t
that co
m
p
letely
ch
ang
e
th
e m
e
an
ing
s
of wo
rds. Detectin
g
its sco
p
e
with
in
a sentence (te
x
t) are n
ecessary in
f
i
n
d
i
ng
out th
e
sentim
ents from
a piece of te
xt.
[27]
Word
sense
di
sambi
g
u
a
ti
on
:
Mo
st v
a
l
u
ab
le con
cep
t t
o
b
e
add
r
essed
wit
h
resp
ect t
o
sen
t
i
m
en
t an
alysis. Th
e non
literal senses,
suc
h
as m
e
t
a
pho
rs a
n
d e
x
pa
nde
d
sense
s
, t
e
nd
t
o
i
n
di
cat
e su
b
j
ect
i
v
i
t
y
, t
r
i
g
geri
n
g
p
o
l
a
r
i
t
y
and e
ffect
i
n
g
t
h
e
resu
lted op
in
i
o
n
su
mm
arizat
i
o
n.[28
]
Dom
a
i
n
ad
ap
t
a
ti
on
p
r
o
b
l
ems:
The
sent
i
m
ent
cl
assi
fi
er t
r
ai
n
e
d
wi
t
h
t
h
e l
a
b
e
l
e
d
dat
a
f
r
om
o
n
e
d
o
m
a
i
n
n
o
rm
al
l
y
cannot
pe
rf
orm
u
p
to
th
e m
a
rk
i
n
an
o
t
h
e
r
domain
.
Th
is
p
r
o
b
l
em
is ter
m
ed
as t
h
e do
main
ad
ap
tatio
n p
r
ob
lem
in
sen
t
i
m
en
t
cl
assi
fi
cat
i
on b
y
usi
ng som
e
label
e
d dat
a
f
r
o
m
t
h
e source d
o
m
a
i
n
and a l
a
rge am
ount
o
f
unl
a
b
el
ed dat
a
fr
om
th
e targ
et do
main
.
[29
]
4.
CO
NCL
USI
O
N
Thi
s
w
o
r
k
p
r
es
ent
s
an i
n
-
d
e
p
t
h
bac
k
g
r
ou
n
d
st
udy
ab
out
o
p
i
ni
on m
i
ni
ng.
The su
b
j
ect
ha
s fasci
n
at
ed
co
nsid
erab
le co
n
c
en
tration
si
n
ce th
e
1
990
s, in
p
a
rticu
l
ar
with res
p
ect t
o
s
u
bjectivity analysis and l
e
xical
reso
u
r
ce ge
ner
a
t
i
on. B
a
se
d o
n
t
h
e s
u
r
v
ey
d
one
we ha
ve s
een t
h
at
i
n
20
1
3
f
o
cu
s was
gi
ven t
o
de
vel
o
p
m
ent
of
m
odel
and fr
am
ewor
ks f
o
r
sent
im
ent
anal
y
s
i
s
and
in
2
0
1
4
it h
a
s sh
ifted
to
co
n
t
en
t ex
tractio
n
and
cl
assi
fi
cat
i
on a
nd t
h
e cu
rre
nt
y
ear foc
u
s i
s
m
o
st
ly
headi
n
g t
o
war
d
s as
pe
ct
based
pre
d
i
c
t
i
on,
whi
c
h c
oul
d be
m
u
ch usef
ul
i
n
t
h
e
sem
a
nt
i
c
we
b a
n
d
com
m
on sense
k
n
o
wl
e
dge
.
A
nu
m
b
er of c
o
m
put
at
i
onal
m
ode
l
s
and
l
i
ngui
st
i
c
feat
ures
rel
a
t
e
d t
o
o
p
i
n
i
o
n m
i
ni
n
g
, com
p
o
n
e
nt
anal
y
s
i
s
and
opi
ni
o
n
-t
a
r
get
i
d
ent
i
f
i
cat
i
on a
r
e
t
h
o
r
o
u
ghl
y
di
scusse
d.
REFERE
NC
ES
[1]
Jack Dorsey
, Tw
itter
Usage, 201
5, Avai
lab
l
e at
h
ttps://
about
.twit
t
e
r.com
/
com
p
an
y.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
ECE
I
S
SN
:
208
8-8
7
0
8
Onl
i
n
e
C
r
ow
ds
O
p
i
n
i
o
n
-
Mi
ni
ng
i
t
t
o
A
n
al
yz
e C
u
rre
nt
Tre
n
d:
A Revi
ew
(
H
ari
t
h
a Akki
nen
i
)
1
186
[2]
A. Tumasjan
,
T.O. Sprenger
,
P.G Sandner,
and
I
.
M. Welp
e. Pred
icting
Elections
with Twitter
:
W
h
at 140
Char
acters
Reveal about P
o
litical Sen
timent. In Internatio
nal
AAAI conference on Weblo
g
s and Social
Media (ICWSM),
2010.
[3]
J. Bollen, H. Mao, and
X.J. Ze
ng
. Twitter Mood
Predicts
the Stoc
k Market. CoRR
, abs/1010
.3003, 2010.
[4]
A. Lam
b
, M.J. Paul, and M. Dre
d
ze.
Separ
a
t
i
ng Fact from
Fear: Track
ing Flu
Infect
ions on Twitter. In Conferen
c
e
of the North American Chapter of the Associatio
n for
Computational Lingu
istics: Hu
man Language Techno
logies
,
pages 789,7
95, June 2013.
[5]
Ionia Ver
itaw
a
ti, I
t
o Wasito
,
T Basarudd
in
,
Text Prepro
cessing using Annotated
Suffix Tr
ee with
Match
i
n
g
Key
phr
ase, Vol
5, No 3
,
In
ternational Journal
of
Electrical and
C
o
m
puter Engin
e
ering, 2015.
[6]
J. Trout and S. Rivkin, Th
e New Pos
itioning: The Latest on the World’s #1B
usi
n
ess Strategy
.
New York: McGr
aw
Hill, 1996.
[7]
Jer
y
Alth
af, Sreevas S, Crowdsourcing Platform
For Approc
hing Humanitarian C
h
allenges,
Intern
ationl Journal of
Social
Networki
ng and Vir
t
ua
l C
o
m
m
unities,Vol1, No2,
Dec
201
2,
[8]
Boundless. "Ty
p
es of Consumer Genera
ted Digital Con
t
ent" Bo
undless MArke
ting, 3, Jul, 201
4, retriev
e
d fro
m
https://www.bou
ndless.com/
marketing/textbo
oks/boundl
ess-marketing-textbook
/
[9]
Ge Xu, Houfeng Wang, 2011. Se
mi-supervised Chinese contex
tual polarit
y
classification with automatic featur
e
selection. Paper
presented
at th
e
2011 Seventh
In
tern
ational Conf
erence on
Na
tur
a
l Computation
(
I
CNC).
[10]
B. Pang and L.
Lee, A sentimen
tal
e
ducation: Sentiment an
aly
s
is
using subjectiv
ity
summarizatio
n based on mini-
mum c
u
t
s
.
[11]
Cambria,
E., H
u
ssain, A., 2012
. Sentic
computing: Techniqu
es
, Tools
,
a
nd Ap
plications, Springer Neth
erlands
,
ISBN: 978-94-007-5069-2.
[12]
Yao W
u
, Mart
in Este
r, FLA
M
E: A Probab
ilisti
c M
odel
Com
b
ining Aspect
Based Opi
n
ion Mining
a
nd
Collabor
ativ
e Filtering
, WSDM’
15, Februar
y
2–
6,
2015, Shang
h
ai, Chin
a. Cop
y
righ
t 2015 ACM 978-1-4503-
3317-7/15/02 h
t
p://dx.do
i.org
/
10
.1145/2684822.2
685291.
[13]
Zheng X, Zvi S, John
H.G, Mu
zaffer U, What can big data and
text anal
y
t
ics tell us about
hotel guestexper
i
ence
and satisfaction
?
, International Journal
of Hospitality
Man
a
gement 44 (2015) 120–130,
http://dx.doi.org/10.1016/j.ijhm.2
014.10.
013
0278
-4319/©
2014 Elsevier Ltd
.
[14]
Wenting Tu, D
a
vid Cheung
, and Nikos Mamoulis, Ti
m
e
-sen
sitive Opinion
Mining for Prediction
,
2015,
Association for
t
h
e Advanc
em
ent
of Artif
ic
ial
Int
e
lligen
ce
[15]
Das, S., Sun, X
.
D.
and Du
tta,
A. (2015). Investigating
User
Riders
hip Sen
t
iments for Bike
Sharing
Programs.
Journal of
Trans
portation
T
echn
o
logies, 5, 69-75
. ht
tp:/
/dx.doi
.or
g
/10.4236/
jtts.20
15.52007
[16]
Thu
y
T.B. Luon
g, Douglas Hous
ton, Publ
ic op
inions of light rail
service in
Lo
s A
ngeles, an
anal
ysis using Twitter
data,
iConferen
ce 2015 Proceed
ings
[17]
Yan Zhao1, Su
y
u
Dong2 and Jin
g
Yang,
Effect
Research of Aspects Extr
act
ion for Chinese Hotel Reviews Based
on Machine
Learning Method
, Intern
ation
a
l
Journal of Smart Home Vol. 9, No. 3 (20
15), pp. 23-34
http://dx.doi.org/10.14257/ijsh.20
15.9.3
.
03
[18]
Chetas
hri Bhad
anea
, Hardi D
a
la
lb, Heen
al
Dos
h
i, S
e
ntim
ent anal
ysis: M
easuring opinio
n
s, Interna
tion
a
l
Conferenc
e
on
Advanced Com
p
uting
Techno
log
i
es and
Appli
cat
i
ons (IC
ACTA-2015), pro
ceed
ing
s
.
[19]
Rahul Tejwan
i, Twodimensional
Sentiment Analy
s
is of
tex
t
, arXiv:1406.2022v1 [
c
s.IR]
8 Jun 201
4
[20]
Nathan A, Timothy
M,
Jacob
L, Garre
t
t
H,
Dan
e
L
,
W
e
i Hu,
Se
ntim
ent Ana
l
y
s
i
s
on the
Socia
l
Networks Using
Stream Algorith
ms, Journal of Data
Analy
s
is and Information Processing,
201
4, 2, 60-66,Published Online M
a
y
2014 in SciRes h
ttp://www.scir
p.org/journal/jd
aip
,
http://dx
.
doi.o
rg
/10.4236/jdaip
.
2
014.22008
[21]
Matheus A, Polly
a
nn
a G, Meey
o
ung C, Fabrício
B, iFee
l: A Web
S
y
stem that Compares and Combines Sentiment
Analy
s
is Metho
d
s, International
World Wide Web Conference Committ
ee (I
W3C2). April 7–11, 2014, Seo
u
l,
Korea. ACM 978-1-4503-2744-2/14/04.
[22]
Linus Philip L
a
wrence
, Rel
i
abi
l
i
t
y
of Sentim
ent
Mining Tool
s: A
com
p
arison of Sem
a
ntria
and Socia
l
Mention
,
3r
d
IBA Bachelor
Thesis Conferen
ce, July
3rd, 2014
, Ensch
e
de,
The
Netherlands.
[23]
Felipe B, Mar
c
e
l
o M, Barbara P
,
Com
b
ining Strengths,
Em
otion
s
and Polarities
for Boosting Twitter Sent
im
ent
Analy
s
is
, WISDOM’13, August
11
2013, Chicag
o, IL, USA Cop
y
righ
t 2013
ACM 978-1-4503-2332-1/13/08.
[24]
Ben He, Craig
Macdonald
, I
a
dh
Ounis,
Rank
ing
Opinionated B
l
og Posts using Op
inionFinder
Cop
y
righ
t is h
e
ld b
y
the
author/own
er
(s). SIGIR’08, J
u
ly
20–24, 2008, Singapore. AC
M 978-1-60558-164-4/08/07.
[25]
Aino R, MaritaV, Social Med
i
a Monitor
i
ng f
o
r Crisis
Communication: Pro
cess, Me
thods
and Tr
ends in
the
Scientif
ic
Li
ter
a
t
u
re,
Online
Journal of
Com
m
u
nica
tion
a
nd Med
i
a Techno
logies
Volume: 4 – Issue: 1 –
Januar
y
-
2014
[26]
Ding, X. and B. Liu. Res
o
lv
ing objec
t and att
r
ibu
t
e coref
e
ren
ce in
opinion m
i
ning.
In P
r
oceedings
of Internat
iona
l
Conference on
C
o
mputationa
l Linguistics (COLI
NG-2010), 2010.
[27]
Councill, I
.
G.,
Mcdonald, R
.
an
d Velikovich
, L.
(2010).
What'
s
great and
what'
s
not: learning
to
classif
y
th
e scop
e
of negation for i
m
p
roved sentiment anal
y
s
is
. In: The Wo
rkshop o
n
Negation and Speculation in Natural Languag
e
P
r
oces
s
i
ng. Upp
s
ala,
S
w
eden.As
s
ociation
for
Co
mputational Lin
guistics, 51-59.
[28]
V. Rentoumi,
George G. Sentimen
t Analy
s
is
of Figurative
Languag
e
using
a Word Sense Disambiguatio
n
Approach. Proccedings of In
tern
ational Confer
ence
RANLP 2009 -
Borovets, Bulg
aria, pag
e
s 370–3
75, 2009
.
[29]
Shoushan Li,
Yunxia Xue, Zhongqing Wang, Guodong Zho
u
, Active
Lear
ning for
Cross-omain Sentiment
Clas
s
i
fic
a
tion
,
P
r
oceed
ings
of
th
e Twen
t
y
-Thi
rd
Interna
tiona
l J
o
i
n
t Confer
enc
e
o
n
Artifi
c
ia
l In
tel
l
i
genc
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
I
J
ECE Vo
l. 5
,
N
o
. 5
,
O
c
tob
e
r
20
15
:
118
0
–
11
87
1
187
BIOGRAP
HI
ES OF
AUTH
ORS
M
r
s
.
Har
i
tha A
kkine
ni
receiv
e
d her M
a
s
t
er of I
n
form
ation Te
ch
nolog
y
d
e
gre
e
fr
om
Univers
i
t
y
of Ballarat
in 2
005.She is pursuing Ph.D (C
SE) from Koneru Lakshmaiah
University
. Her
inter
e
s
t
ed a
r
eas
are Dat
a
M
i
ning
, Im
age M
i
ning
, Opnion Mining, Sentimen
t an
al
ys
is
and S
o
cial
Networks. Published papers
in v
a
rious repu
te
d
In
terna
tiona
l Journ
a
ls
in
cluding
Springer.
Dr.
B.
Vijay
a
Babu
, B.Tech, M
.
Tech
, P
h
.D is
working as
a profes
s
o
r at KL Uni
v
ers
i
t
y
, He has
got his Ph.D fro
m Andhra University
in
y
e
ar 20
12,
He h
a
s got 2
0
y
ears of
teach
ing exper
i
ence.
His
area of res
earch is
Dat
a
M
i
ning/Knowle
dge Engineerin
g and published more than 25
research
publications in v
a
rio
u
s National an
d Internation
a
l Journals, in
cluding SCOPUS
indexed
.
Dr
P
.
V.
S Lakshmi
, B.Tech
, M
S
, Ph.D working as professor i
n
PVP Siddhartha Institu
te of
Techno
log
y
Vi
j
a
yawada
. S
h
e has
got 21
y
e
ars
of teach
ing exp
e
rien
ce and 6
years
of res
ear
ch
experience. She was awarded with Ph.D in the
y
e
ar 2011in the ar
ea of Arificial Intelligen
ce and
Image Processing. She has published papers in va
rious reputed journals includ
in
g Springer.Her
areas
of
int
e
res
t
s
are Da
ta
M
i
ni
ng, Dat
a
bas
e
M
a
nagem
e
nt
S
y
s
t
em
s
,
Im
age P
r
o
ces
s
i
ng, S
o
c
i
al
Networks and S
e
ntiment Analy
s
is.
Evaluation Warning : The document was created with Spire.PDF for Python.