Inter
national
J
our
nal
of
Inf
ormatics
and
Communication
T
echnology
(IJ-ICT)
V
ol.
14,
No.
1,
April
2025,
pp.
1
∼
10
ISSN:
2252-8776,
DOI:
10.11591/ijict.v14i1.pp1-10
❒
1
A
sur
v
ey
on
no
v
el
appr
oach
to
semantic
computing
f
or
domain
specic
multi-lingual
man-machine
interaction
Anjali
Bohra,
Nemi
Chand
Barwar
Department
of
Computer
Sciences
and
Engineering,
MBM
Uni
v
ersity
,
Jodhpur
,
India
Article
Inf
o
Article
history:
Recei
v
ed
Mar
5,
2024
Re
vised
Jul
24,
2024
Accepted
Sep
21,
2024
K
eyw
ords:
Deep
learning
Karaka
relations
Machine
learning
Natural
language
processing
P
anini
grammar
Semantic
computing
Semantic
role
labeling
ABSTRA
CT
Natural
language
processing
(NLP)
helps
computational
linguists
to
understand,
process,
and
e
xtract
information
from
natural
languages.
Linguist
P
anini
signi-
es
’information
coding’
in
a
language
and
e
xplains
that
Karakas
are
semantico-
syntactic
relations
between
nouns
and
v
erbs
that
resemble
participant
roles
of
modern
case
grammar
.
Computational
grammar
maps
vibhakti
(inections)
of
nominals
and
v
erbs
to
their
participant
roles.
Karaka’
s
theory
e
xtracts
semantic
roles
in
the
sentences
which
act
as
intermediate
steps
for
v
arious
NLP
tasks.
The
surv
e
y
sho
ws
that
NLP
seeks
to
bri
dge
the
g
ap
for
man-machine
interaction.
The
w
ork
presents
the
impact
of
machine
learning
on
natural
language
processing
with
changing
trends
from
traditional
to
modern
scenarios
with
P
anini’
s
classi-
cation
scheme
for
semantic
computing
f
acilitating
machine
understanding.
The
study
presents
the
signicance
of
Karaka
for
semantic
computing,
methodolo-
gies
for
e
xtracting
semantic
roles,
and
analysis
of
v
arious
deep
learning-based
language
processing
systems
for
applications
lik
e
question
answering.
The
sur
-
v
e
y
co
v
ered
around
50
research
articles
and
21
Karaka-based
NLP
systems
per
-
forming
multiple
tasks
lik
e
machine
translation,
question-a
nswering
systems,
and
te
xt
summaries
using
machine
learning
tools
and
frame
w
orks.
The
w
ork
includes
surv
e
ys
from
reno
wned
journals,
books,
and
rele
v
ant
conferences,
as
well
as
descriptions
of
the
latest
trends
and
technologies
in
the
machine
learning
domain.
This
is
an
open
access
article
under
the
CC
BY
-SA
license
.
Corresponding
A
uthor:
Anjali
Bohra
Department
of
Computer
Science
and
Engineering,
MBM
Uni
v
ersity
Jodhpur
,
Rajasthan,
India
Email:
anjali
vb
.phdcse@mbm.ac.in
1.
INTR
ODUCTION
Articial
intelligence
(AI)
inculcates
human
abilities
into
machines
by
allo
wing
learning
through
e
x-
perience
and
adjusting
to
ne
w
inputs.
Examples
include
computers
playing
cards,
and
digital
assistants
lik
e
Siri
[1].
Computers
are
trained
using
AI
technologies
including
machine
learning,
natural
language
processing
(NLP),
and
computer
vision
to
accomplish
specic
tasks
[2].
Machines
are
trained
through
machine
learn-
ing
algorithms
using
data
analysis
and
a
v
ailable
patterns
with
minimal
human
interv
entions.
Computers
can
communicate
with
humans
in
their
language,
through
reading,
identifying
and
classifying
te
xt,
hearing
and
interpreting
speech,
and
measuring
sentiments
using
NLP
techniques.
Computer
vision
trains
computers
to
analyze
and
understand
the
visual
w
orld
by
accurately
identifying,
and
classifying
objects,
recognizing
f
aces,
processing
li
v
e
ac
tions
of
a
football
g
ame,
and
surpassing
human
visual
abilities
in
man
y
areas.
Free
multi-
lingual
machine
translators
de
v
eloped
by
Google
and
Ale
xa
de
v
eloped
by
Amazon
are
prominent
e
xamples.
AI
J
ournal
homepage:
http://ijict.iaescor
e
.com
Evaluation Warning : The document was created with Spire.PDF for Python.
2
❒
ISSN:
2252-8776
technologies
ha
v
e
transformed
communication
technology
by
shifting
the
data-dri
v
en
paradigm
to
intelligence-
dri
v
en
endea
v
ors.
NLP
helps
machines
to
unders
tand
human
language
and
beha
v
e
as
inte
lligently
as
humans
by
amalg
amation
of
linguistics
and
computer
science
disciplines
[3].
NLP
analyzes
dif
ferent
aspects
of
language
lik
e
syntax,
semantics,
pragmatics,
and
morphology
to
transform
linguistic
kno
wledge
into
production-based
algorithms
for
problem
solution
[4],
[5].
T
asks
include
translation,
relationship
e
xtraction,
speech
recognition,
named-entity
recognition,
topic
se
gmentation,
sentiment
analysis,
Chatbots,
and
te
xt
Summarization.
NLP
tasks
are
performed
in
a
sequence
using
a
corpus
and
frame
w
ork.
A
frame
w
ork
denes
learning
models
us-
ing
components
that
automatically
understand,
code,
compute
gradients,
and
perform
parallel
processing
for
optimized
performance
[6].
Basic
approaches
to
NLP
are
distrib
utional-based,
frame-based,
model-theoretical-based,
and
interacti
v
e-
based
learning
[7].
Distrib
utional-based
approaches
use
statistical
concepts
focused
on
mathematical
analysis
of
the
content,
including
tasks
lik
e
part-of-speech
tagging,
dependenc
y
parsing,
and
semantic
relationships.
Frame-based
approaches
consider
f
rames
as
the
standard
for
representing
concepts
.
Model-theoretical-based
approaches
are
semantic
methods
where
the
model
denes
the
idea
related
to
the
concept
and
meaning
of
the
sentence.
Interacti
v
e
learning
approaches
consider
pragmatic
concepts.
T
able
1
sho
ws
a
list
of
designed
lan-
guage
processing
systems
lik
e
sentiment
analyzer
,
part
of
speech
tagger
,
and
emotion
detection
system
through
NLP
methods
and
approaches.
Understanding
natural
language
has
three
stages
of
de
v
elopment:
the
rationalist
stage,
the
empirical
stage,
and
the
deep
learning
stage.
T
able
1.
Lang
auge
processing
systems
with
NLP
methods
S
No
NLP
systems
NLP
methods
with
approaches
1
Sentiment
analyser
[8]
T
opic
as
features
(distrib
utional
approach)
2
P
arts
of
speech
taggers
[9]
Rule
based
methods
(distrib
utional
approach)
3
Chunking
[10],
[11]
Log-linear
method/multi-label
classication
4
Named
entity
recognition
system
[12],
[13]
Statistical
methods
(distrib
utional
approach)
5
Emotion
detection
system
[14]
Conditional
random
eld
method
(model-theoretical
approach)
6
Semantic
role
labelling
system
[15]
Semantic
representation
(frame-based-approach)
7
Ev
ent
disco
v
ery
system
[16]
Latent
semantic
method
(distrib
utional
approach)
The
rationalist
stage
focuses
on
implementing
Chomsk
y’
s
rules
for
inducing
reasoning
and
kno
wl-
edge
into
NLP
systems
lik
e
ELIZA,
and
MARGIE.
The
empirical
stage
focused
on
implementing
generalized
concepts
in
machines
through
pattern
recognition
and
gener
ati
v
e
models
lik
e
HMM
and
IBM
translation
mod-
els.
The
curre
nt
deep
learning
stage
focuses
on
implementing
a
layered
model
to
perform
end-to-end
learning
for
feature
e
xtraction.
Dense
representations
of
w
ords,
sentences,
paragraphs,
and
documents
are
learned
to
capture
both
syntactic
and
semantic
features.
The
numbers
in
w
ord
v
ector
representation
sho
w
the
closeness
of
the
encoded
meaning
with
the
specied
concept
[17].
NLP
applications
are
hard
and
challenging
as
program-
ming
languages
lik
e
Ja
v
a
and
Python
are
required
for
man-machine
interaction.
These
programming
languages
are
structured
and
unambiguous
while
human
languages
are
ambiguous
as
well
as
re
gion
adapti
v
e
[18].
The
most
dif
cult
part
of
training
computers
using
programming
languages
is
handling
le
xical,
referential,
and
syntax-le
v
el
ambiguity
with
synon
yms
and
h
ypern
yms.
Semantic
computing
concentrates
on
understanding
the
meaning,
interpretation,
and
relationships
be-
tween
w
ords,
phras
es,
and
sentences
through
the
grammar
of
a
language
to
bridge
the
g
ap
between
people
and
computers
[19].
It
composes
information
based
on
meaning
and
v
ocab
ulary
by
implementing
computing
technologies
(lik
e
articial
intelligence)
through
NLP
,
kno
wledge
engineering,
softw
are
engineering,
and
com-
puter
netw
orks
to
e
xtract,
transform,
and
synthesize
the
content
[20]–[22].
The
k
e
y
components
of
semantic
analysis
are
le
xical
semantics,
syntax
w
ord
embedding,
and
v
ector
space
models.
The
study
in
v
estig
ates
the
ef
fect
of
deep
learning
for
NLP
which
has
achie
v
ed
ne
w
benchmarks
through
distrib
uted
representation
and
se-
mantic
generalization
of
w
ords.
Conte
xtual
w
ord
embeddings
in
dif
ferent
conte
xts
sho
w
dif
ferent
real-v
alued
v
ector
representations
for
the
same
w
ord
from
a
corpus
[23],
[24].
W
ord
embedding
of
te
xtua
l
data
is
obtained
using
the
embedding
layer
of
K
eras
deep
learning
frame
w
ork,
W
ord2V
ec
or
GloV
e
model,
and
bidirectional
encoder
representations
from
transformers
(BER
T)
language
model
[25],
[26].
Pre-trained
embeddings
ha
v
e
sho
wn
remarkable
impro
v
ement
in
NLP
tasks
lik
e
speech
recognition,
syntactic
parsing,
te
xt
understanding
and
summarization,
and
question-answering
systems
[27]–[29].
Challenges
in
NLP:
Despite
major
success
in
v
arious
NLP
tasks
lik
e
language
modeling
and
machine
Int
J
Inf
&
Commun
T
echnol,
V
ol.
14,
No.
1,
April
2025:
1–10
Evaluation Warning : The document was created with Spire.PDF for Python.
Int
J
Inf
&
Commun
T
echnol
ISSN:
2252-8776
❒
3
translation,
deep
l
earning
methods
persist
in
lack
of
interpretability
to
interpret
inter
-sentential
relations.
More
w
ork
is
required
in
neural-symbolic
representation
of
human
kno
wledge
[30].
Decienc
y
of
kno
wledge,
in-
terpretability
of
models,
and
requirement
of
lar
ge
datasets
are
the
major
challenges
for
NLP
through
deep
learning.
Reinforcement
learning
with
inference,
and
kno
wledge-base
lead
to
ne
w
learning
paradigms
[31].
Pragmatic
interpretation
is
still
an
open
area
of
research
[32].
W
ord
sense
disambiguation,
structural
ambigu-
ity
,
and
co-reference
resolution
are
challenges
due
to
ambiguity
and
polysemy
.
Idiomatic
e
xpress
ions
require
conte
xtual
or
cultural
understanding.
Lack
of
domain-speci
c
kno
wledge
misinterprets
sentential
relationships
because
dif
ferent
re
gions
include
unique
terms
and
jar
gon
that
are
unf
amiliar
to
generalized
language
process-
ing
systems.
Research
g
aps
in
this
paper
include:
−
Challenge
is
to
de
v
elop
a
uni
v
ersal
approach,
lar
ge
language
model
(LLM)
based
on
Karak
relations
for
mainstream
Sanskrit,
Hindi,
and
English
language
for
NLP
tasks
lik
e
summarization,
and
translation.
−
LLM
suf
fers
from
hallucination
which
can
be
resolv
ed
through
e
xact
topic
e
xtraction
techniques
using
semantic
processing.
−
The
wide
scope
of
research
is
open
for
multi-modal
LLMs
that
combine
te
xt
processing
with
audio,
image,
and
video.
W
ork
outline
in
this
paper
is
brings
together
researchers
fr
om
disciplines
such
as
NLP
,
multi
media
semantics,
semantic
W
eb,
and
pattern
recognition
to
pro
vide
a
single
source
for
presenting
the
state
of
the
technology
to
breakthroughs
on
the
horizon.
The
introduction
co
v
ers
the
history
and
de
v
elopment
of
machine
learning’
s
rele
v
ance
to
natural
language
processing
with
chal
lenges
to
the
eld.
Section
2
co
v
ers
the
back-
ground
for
NLP
in
semantic
processing
with
the
signicance
of
Karak
theory
.
The
ne
xt
section
e
xplains
the
methods
follo
wed
with
result
s
and
discussion.
The
last
section
summarizes
the
w
ork
wit
h
guidelines
for
future
directions.
2.
B
A
CKGR
OUND:
COMPONENTS
OF
A
LANGU
A
GE
Linguistics
considers
language
as
a
group
of
arbitrary
v
ocal
signs,
go
v
erned
by
innate
and
uni
v
ersal
rules
(grammar)
of
the
language.
Grammar
has
tw
o
types:
descripti
v
e
and
Perspecti
v
e
grammar
.
Descripti
v
e
grammar
denes
a
set
of
rules
to
formulate
the
speak
er’
s
grammar
.
Perspecti
v
e
grammar
focuses
on
correctness
in
the
language.
A
grammatical
cate
gory
is
a
class
of
units
or
features
of
a
language
indicating
number
,
gender
,
de
gree,
person,
case,
deniteness
or
indeniteness,
tense,
aspect,
mood,
and
agreement.
Number
is
related
to
singular
or
plural
concepts
while
gender
is
e
xpres
sed
by
v
ariation
in
personal
pronouns
or
third
person.
Examples
of
grammatical
genders
are
he,
she,
it
(singular),
I,
we,
and
you
(rst
a
n
d
second
form),
and
the
y
(third
person
plural
either
common/neuter
gender).
Case
sho
ws
the
relationship
of
the
noun
phrase
with
v
erb
and
other
noun
phrases
in
a
sentence
lik
e
nominati
v
e
case,
geniti
v
e
case,
objecti
v
e
case,
etc,
and
de
gree
is
sho
wn
by
adjecti
v
es
and
adv
erbs.
T
ense
grammatical
cate
gory
represents
a
time
of
an
action
and
aspect
denes
a
vie
w
of
an
e
v
ent
which
can
be
perspecti
v
e
or
imperati
v
e.
Mood
sho
ws
the
speak
er’
s
attitude
to
w
ards
what
he
or
she
is
talking
about.
Representing
grammar
(of
a
language)
as
mathematical
e
xpression
is
an
intractable
problem.
Semantic
netw
orks,
rst-order
logic,
frames,
and
production
systems
are
used
for
kno
wledge
representation.
Semantic
netw
orks
describe
the
relation
between
an
object
and
a
class.
Prolog
programming
language
is
based
on
a
subset
of
rst-order
l
o
gi
c
which
is
a
declarati
v
e
language
for
writing
logic
stat
ements
and
proofs.
The
kno
wledge
is
con
v
erted
into
modular
chunks
using
a
frame-base
approach
while
rules
specifying
patterns
and
actions
are
specied
through
a
production
system-based
approach.
2.1.
Karak
theory
In
linguistics,
semantic
analysis
represents
syntactic
structures
(w
ords
and
phrases)
with
their
language-
independent
meaning
[33].
Linguist
P
anini
dened
a
Karak-based
approach
for
te
xt
and
speech
processing.
He
dened
kno
wledge
representation
methodologies
in
his
book
‘
Asthadhayayi’
which
are
equi
v
alent
to
current
AI
systems
including
meta-rules
for
coding
AI
softw
are
[34].
He
de
v
eloped
a
frame
w
ork
for
uni
v
ersal
grammar
that
can
be
applied
to
an
y
natural
language
[35]–[37].
The
frame
w
ork
is
based
on
the
concept
of
karma
and
morphosyntactic
structures
to
e
xtract
semantic
roles
in
a
sentence.
A
semantic
role
describes
the
relation
of
a
syntactic
constituent
(noun
phrase)
with
a
predicate
(the
v
erb
or
action)
as
an
agent,
patient,
and
instrument
[38].
P
aninian
grammar
processes
sent
ences
at
four
le
v
els
namely
surf
ace
le
v
el
(uttered
sentence),
bhakti
le
v
el,
Karaka
le
v
el,
and
semantic
le
v
el.
Karakas
specify
relations
between
nominal
and
v
erbal
root
[39].
F
ollo
wing
A
surve
y
on
no
vel
appr
oac
h
to
semantic
computing
for
domain
specic
.
.
.
(Anjali
Bohr
a)
Evaluation Warning : The document was created with Spire.PDF for Python.
4
❒
ISSN:
2252-8776
are
the
six
Karakas
specied
by
P
anini
according
to
their
participation
with
the
v
erb
in
a
sentence:
i)
Karta:
describes
action
of
v
erb;
ii)
Karam:
desired
by
the
Karta
Karak
(subject);
iii)
Karana:
act
as
instrum
ent
of
the
action
performed
by
Karta;
i
v)
Sampradaan:
act
as
recipient
of
an
action;
v)
Apaadaan:
e
xpress
detachment
or
comparision
from
a
source;
and
vi)
Adhikarana:
describe
place
of
action.
The
Karaka-based
approach
is
a
template-based
generation
system
which
answer
Karak-based
ques-
tions
with
rele
v
ance
to
the
case
of
noun
phrases
in
the
sentence.
A
noun
or
pronoun
e
xists
in
eight
forms
in
a
sentence
and
therefore
causes
eight
types
of
cases.
Se
v
en
forms
of
vibhakti
are
nominati
v
e,
accusati
v
e,
instrument,
dati
v
e,
locati
v
e,
genti
v
e,
and
v
ocati
v
e
[40].
Karaka
relations
are
semantic-syntactic
relations
where
Karta
Karak
acts
as
a
nominati
v
e
case,
Karam
Karak
as
an
objecti
v
e/accusati
v
e
case,
Karan
Karak
as
an
in-
strument,
Sampradaan
as
a
dati
v
e
case,
Apadan
as
an
Ablati
v
e
case
and
Sambandh
is
geneti
v
e/possessi
v
e
case.
Adhikaran
Karak
act
as
a
locati
v
e
case
and
Sambodhan
as
a
v
ocati
v
e
case
[41].
Case
is
a
property
shared
by
all
the
languages
of
the
w
orld
[42].
2.2.
Methods
of
semantic
pr
ocessing
Semantic
processing
focuses
on
w
ords
to
determine
their
signicance
in
a
phrase
or
a
sentence
[43].
Similarity
measures
are
used
to
nd
the
rele
v
anc
y
between
the
w
ords
[44].
Semantic
processing
methods
decode
the
meaning
within
the
te
xt.
The
process
starts
with
preprocessing
and
le
xical
analysis
follo
wed
by
parsing
and
syntactic
analysis,
semantic
frame
identication,
and
establishing
mathematical
representation
of
w
ords
through
v
ector
space
models/embedding
l
ayers.
Based
on
the
required
application
suitabl
e
semantic
analysis
method
i
s
selected
to
e
xtract
the
features.
Finally
,
the
system
is
e
v
aluated
for
impro
ving
the
per
-
formance
using
techniques
such
as
semantic
feature
analys
is,
latent
semantic
analysis,
and
semantic
content
analysis
[45].
−
Semantic
feature
analysis
emphasizes
the
representation
of
w
ord
features
through
feature
selection
(part
of
speech
(POS)
and
morphological
features),
determining
weights
(through
term
frequenc
y
,
in
v
erse-term
frequenc
y
,
normalized
term
frequenc
y
,
and
global
term
weighting),
and
similarity
measurement(through
cosine/Jaccard
similarity
and
euclidean
distance).
−
Latent
semantic
analysis
captures
the
relationship
of
w
ords
with
their
conte
xt
using
statistical
methods
lik
e
reducing
dimensionality
and
comparing
semantic
similarity
.
It
is
the
mathematical
method
for
e
xtracting
the
meaning
of
w
ords.
The
mathematics
is
to
obtain
parameters
of
an
y
X
rectangular
tXp
matrix
of
(r
rank)
terms
and
passage
through
decomposition
into
three
matrices
using
singular
v
alue
decomposition
using
(1).
X
=
T
S
P
T
(1)
where
T
is
txr
matrix
with
orthonormal
columns,
P
is
pxr
matrix
with
orthonormal
columns
and
S
is
r
x
r
diagonal
matrix
with
sorted
entries
in
descending
order
[46].
−
Semantic
content
analysis
identies
relationships
between
w
ords
and
phrases
using
dependenc
y
parsing
(graph-based
parsing),
thematic
roles
and
case
roles
(re
v
eals
relationships
between
actions,
participants,
and
objects),
and
identication
of
semantic
frame.
3.
LITERA
TURE
REVIEW
Anusaarka,
a
language
translation
system
based
on
paninian
theory
uses
an
interlingua-based
ap-
proach
which
is
an
intermediate
representation
dened
by
v
erb,
noun,
and
Karaka
relations
[47]–[49].
A
rule-based
Hindi
lemmatizer
that
generates
the
rules
for
e
xtracting
suf
x
es
from
the
gi
v
en
w
ord
[50],
[51].
The
go
v
ernment
of
India
proposed
a
supervised
learning-based
Beng
ali
root
w
ord
e
xtraction
system
using
P
aninian
grammatical
rules
under
the
TDIL
project
[52].
Opinion
classication
system
for
Odia
language
us-
ing
syntactic-semantic
concept
[53].
A
list
of
dependenc
y
relations
w
as
prepared
based
on
P
anini’
s
grammar
which
sho
ws
that
relations
represent
well-dened
semantics
for
e
xtraction
from
the
surf
ace
form
of
the
w
ord
without
an
y
linguistic
information
[54].
Designed
a
paninian
frame
w
ork-based
case
mark
er
error
-resolv
er
for
Indian
languages
[55].
A
Marathi
T
reebank
w
as
also
designed
based
on
Karak
theory
using
Marathi
corpus
[56].
Natural
language
interf
ace
for
dat
abases
w
as
designed
to
process
user
queries(including
logical
operators,
relational
operators,
and
joining
of
tables
for
the
Hindi
language)
by
con
v
erting
them
into
equi
v
alent
standard
structured
query
language
(SQL)
query
through
computational
P
aninian
grammatical
frame
w
ork
[57].
De-
signed
a
constraint-based
P
arser
for
the
Nepali
language
using
Karak
theory
[58].
Int
J
Inf
&
Commun
T
echnol,
V
ol.
14,
No.
1,
April
2025:
1–10
Evaluation Warning : The document was created with Spire.PDF for Python.
Int
J
Inf
&
Commun
T
echnol
ISSN:
2252-8776
❒
5
T
able
2
summarizes
Karak-based
language
processing
systems
performing
machine
translation
tas
ks
for
Hindi,
Sanskrit,
and
Malayalam
languages,
parsing
of
languages,
and
question-answering
systems.
The
de-
scription
includes
their
functioning,
used
methodology
,
datasets
or
corpus
as
well
as
e
v
aluation
results.
These
systems
use
denite
w
ords
or
sentences
from
specic
corpus
or
datasets
which
are
trained
with
features
ob-
tained
from
semantic
processing.
The
systems
are
e
v
aluated
using
preci
sion
and
recall
F-measure.
All
systems
attained
almost
75
to
95
percent
accurac
y
in
results.
T
able
2.
Karka-based
language
processing
systems
S
No.
System
name
Description
Language
Method
Accurac
y
Corpus/Dataset
1
Anusaarka
A
language
trans-
lation
system
Kannada
to
Hindi,
marathi,
Beng
ali,
and
T
elugu
Interlingua
based
method
92%
approx.
30,000
w
ords
from
Kannada
dictionary
and
other
language
dictionaries
2
Hindi
Lem-
matizer
[59]
Extracts
suf
x
es
from
the
root
w
ord
Hindi
P
aradigm
based
method
0.89
2,500
w
ords
for
Hindi
dictionary
3
Root
w
ord
e
xtraction
system
Extracts
Beng
ali
root
w
ord
Beng
ali
Rule
based
method
0.99
10,000
dif
ferent
in-
ected
w
ords
from
Beng
ali
dictionary
4
Opinion
classication
system
[60]
Classies
opinion
of
re
vie
wers
Beng
ali
T
opic
based
ap-
proach
0.7
Beng
ali
ne
wspa-
per
a
v
ailable
at
http://www
.ananda
bazar
.com/
5
Dependenc
y-
relations
identication
system
[61]
lists
dependenc
y
relations
Sanskrit
Production-based
system
0.9
Bhagv
at-Gita
6
Case-
mark
er
-
errors
iden-
tication
system
[62]
Identies
case
mark
er
errors
for
Indian
languages
committed
by
google
machine
translators
English
to
Urdu
translation
Karak-vibhakti
based
dependenc
y
frame
w
ork
Machine
trans-
lation
neural
based
32%
ac-
curate
and
21%
phrase-based
500
English
sen-
tences
7
Sanskrit
Karak
ana-
lyzer
[13]
T
ak
es
unicode
De
vnagri
te
xt
and
returns
Karak
analyzed
te
xt
Sanskrit
Rule
based
ap-
proach
84%
accurate
31
Karaka,
72
vib-
hakti
from
sanskrit
dictionary
8
Pilagiarism
detection
system
[63]
Plagiarism
detec-
tion
system
based
on
paninian
frame
w
ork
Malayalam
Machine
learning
approach
Online
Malayalam
ne
wspapers
9
V
erbframator
[64]
Extracts
v
erb
frames
for
the
gi
v
en
sentences
Marathi
Karaka
based
ma-
chine
learning
Generate
v
erb
frames
b
ut
require
some
human
inter
-
v
ention
40,000
Marathi
v
erbs
from
W
ord-
Net
(subset
of
Indo-W
ordNet)
10
Question-
answering
system
[65]
Generates
ques-
tions
in
Hindi
language
Hindi
Karak-based
ma-
chine
learning
5
pt
Lik
ert
scale:
3.019,
3,336
syntactic
and
semantic
mean
30
sentences
from
Hindi
corpus
11
Question
an-
swering
sys-
tem
[66]
Generate
answers
by
comparing
vibhakthi
and
POS
tags
of
question
w
ords
Malayalam
V
ibhathi
and
POS
tagging
based
ap-
proach
Generate
w
ord
le
v
el
answers
Malayalam
corpus
12
Semantic
tagger
and
Karaka
ana-
lyzer
[51]
Perform
tagging
and
identify
Karaka
Hindi
rule-based
ap-
proach
84%
precise
Hindi
corpus
A
surve
y
on
no
vel
appr
oac
h
to
semantic
computing
for
domain
specic
.
.
.
(Anjali
Bohr
a)
Evaluation Warning : The document was created with Spire.PDF for Python.
6
❒
ISSN:
2252-8776
T
able
2.
Karka-based
language
processing
systems
(Continued)
S
No.
System
name
Description
Language
Method
Accurac
y
Corpus/Dataset
13
T
e
xt
Cluster
-
ing
for
a
doc-
ument
[67]
Generate
mean-
ingful
labels
of
the
clusters
Punjabi
Karaka
based
ma-
chine
learning
ap-
proach
95%
precise
Punjabi
corpus
14
Generate
se-
mantic
roles
[68],
[69]
Generic
labels
for
the
tok
ens
of
te
xt
Malayalam
Karaka
based
ma-
chine
learning
ap-
proach
Malayalam
corpus
15
Karakacross:
sentiment
analysis
[70]
Extract
senti-
ments
related
semantic
roles
Dif
ferent
lan-
guages
Sentiment
e
xtrac-
tion
using
Karaka
theory
Multi-lingual
datasets
16
T
e
xt
summa-
rization
sys-
tem
[71]
Perform
single-
document
sum-
marization
Malayalam
SRL
based
on
Karaka
theory
80
%
precise
Online
Malayalam
repository
17
Cross-
lingual
study
based
on
Karaka
[72]
Impact
of
Karakas
on
congition
Sanskrit,
Marathi,
Kanada,
and
T
elugu
Karaka
based
ma-
chine
learning
sys-
tem
Karta
and
Karma
mapped
accurately
Sanskrit
and
Marathi
language
corpus
18
Case
ana-
lyzer
system
[73]
Extract
cases
of
Eastern
Indo-
Aryan
languages
7
Indo-Aryan
lan-
guages
T
radition
and
modern
approach
to
study
cogniti
v
e
frame
w
ork
80%
accurate
language-
specic
case
relations
Corpus
of
Indo-
Aryan
languages
19
Question
an-
swering
sys-
tem
[74]
Extraction
of
similarity
features
for
classication
in
question
answer
(QA)
selection
Hindi
Karaka
based
ma-
chine
learning
ap-
proach
Proper
e
x-
traction
of
Karaka
reduce
needs
role
of
pre-trained
Hindi
corpus
20
T
e
xt
summa-
rization
sys-
tem
[75]
Extracti
v
e
sum-
marization
of
a
document
Malayalam
Machine
learning
based
on
Karaka
theory
66%
precise
and
65%
ef
-
cient
in
recall
Malayalam
corpus
21
Question
an-
swering
sys-
tem
[76]
Retrie
v
al
of
an-
swers
for
ques-
tion
answering
Hindi
and
Marathi
QA
based
on
Karaka
theory
for
Indic
languages
80%,
60%
pre-
cise
for
Hindi
and
marathi
language
Hindi
and
Marathi
corpus
4.
METHOD
This
re
vie
w
research
on
Karak-based
multi-lingual
language
processing
systems
is
rele
v
ant
to
answer
questions
related
to
the
semantic
interpretation
of
a
language.
Systemat
ic
literature
re
vie
w
(SLR)
has
three
parts:
planning,
construction,
and
reporting
phase.
The
planning
phase
focuses
on
the
need
for
a
re
vie
w
accompanied
by
research
questions.
The
construction
phase
selects
primary
studies
and
e
xtracts
data
from
those
studies
and
the
nal
stage
disseminates
results.
The
w
ork
e
xplains
the
ef
fecti
v
eness
of
NLP
in
semantics
to
f
acilitate
high-le
v
el
programming
languages
(Prolog
and
Python)
for
computers.
5.
RESUL
TS
AND
DISCUSSION
5.1.
Results:
Karak-based
semantic
rules
in
moder
n
generati
v
e
grammar
Karaka’
s
theory
is
syntact
ic
to
the
semantic
formalization
of
language
aspects.
Case
grammar
de-
scribed
by
l
lmore
re
generated
the
P
aninian
proposal
in
a
modern
linguistic
conte
xt.
He
h
ypothesized
human
equi
v
alent
uni
v
ersal
concepts
for
making
judgments
about
the
e
v
ents
or
actions
using
the
follo
wing
answers
to
raised
5W
(who/what/when/where/wh
y)
based
questions
[77],
[78].
−
Who
is
the
initiator
of
the
action?:
Agent
−
What
is
in
v
olv
ed
in
the
action?:
Instrument
(in
v
olv
ed
object)
−
Who
emphasis
on
the
ef
fect
of
the
action?:
Dati
v
e
−
What
is
the
result
of
the
action?:
F
actiti
v
e
(object)
−
When
and
Where
the
e
v
ent
(or
action)
is
oriented?:
Locati
v
e
−
Wh
y
the
things
are
af
fected
by
the
action?:
Objecti
v
e
P
aninian-based
Karak
species
answers
to
the
questions
for
semantic
interpretation
of
an
y
natural
Int
J
Inf
&
Commun
T
echnol,
V
ol.
14,
No.
1,
April
2025:
1–10
Evaluation Warning : The document was created with Spire.PDF for Python.
Int
J
Inf
&
Commun
T
echnol
ISSN:
2252-8776
❒
7
language.
Le
xical,
morphological,
and
syntactic
features
describe
an
y
language
[79].
Le
xico-syntactic
fea-
tures
include
POS
tagging,
morphological
tagging
includes
root
w
ord,
gender
,
number
,
person,
and
case,
and
syntactic
features
include
head
noun,
chunk
label,
and
dependenc
y
relation.
Semantic
role
labeling
is
a
se-
mantic
parsing
technique
widely
used
in
question-answering
systems
or
information
e
xtraction
systems
that
assign
semantic
roles
to
syntactic
constituents
(ar
guments
of
predicate
in
a
sentence).
Karakas
e
xplained
pars-
ing
Indian
languages
and
creating
T
reebank
for
Hindi
[80].
The
T
reebank
dataset
contains
around
four
mil-
lion
annotated
w
ords
di
vided
into
dif
ferent
annotations
lik
e
parts-of-speech,
syntactic,
and
semanti
c
sk
eletons
[81].
Sanchay
is
a
free
linguistic
annotation
tool
for
Indian
languages
published
in
a
list
of
programs
as
part
of
education.
Dependenc
y-based
formalis
m
is
incorporated
for
morphologically
rich
languages
ef
forts
ha
v
e
been
incorporated
for
dependenc
y-based
formalism
[82],
[83].
Hyderabad
dependenc
y
treebank
(HyDT)
for
Hindi
uses
Karak
relations
to
capture
local
semantics
and
labels
rele
v
ant
to
the
v
erb
through
dependenc
y-based
approach
[84],
[85].
5.2.
Discussion:
method
of
inf
ormation
coding
in
a
language
Language
has
grammar
(rules)
for
combining
the
w
ords
[86].
Languages
use
parsing
to
code
the
information.
Semantic
analysis
helps
in
encoding
the
relations
in
a
sentence.
Grammar
decides
ho
w
the
relations
are
coded
in
the
language.
Information
can
be
summarized
by
answering
5W
questions
lik
e
who,
what,
when,
where,
and
wh
y
.
In
machine
translation,
a
gi
v
en
source
is
translated
into
the
tar
get
language
through
5Ws
comprehensi
v
e
[87].
Answering
5Ws
generates
domain-independent
generic
semantic
roles.
P
aninian
grammar
signies
the
minute
obse
rv
ations
re
g
arding
i
nformation
coding
in
a
language.
P
anini
signies
information
coding’
in
a
language
by
answering
three
questions:
where,
which
and
ho
w
.
Three
aspects
of
questioning
for
e
xtracting
information
coding
in
a
language
are:
Where
the
information
is
coded?
Which
relations
are
coded
in
the
sentence?
And
Ho
w
the
relations
are
coded?
A
w
ord
can
be
tagged
as
nominal/v
erbal
form
according
to
the
grammar
.
T
ense
and
person
morpho-
logically
inect
the
w
ord
in
a
sentence.
Each
sentence
is
represented
using
alphabet
letters
and
one
sentence
can
be
dened
in
terms
of
another
e
xactly
lik
e
the
production
rules
of
a
Chomsk
y
gram
mar
[88].
Surf
ace
le
v
el
(uttered
sentence),
vibhakti
le
v
el,
Karaka
le
v
el,
and
semantic
le
v
el
are
the
four
le
v
els
of
te
xt
processing
using
the
P
aninian
frame
w
ork.
6.
CONCLUSION
The
paper
presents
a
surv
e
y
on
paninia
n
frame
w
ork-based
(Karak
theory-based)
language
proce
ss-
ing
systems.
It
deals
with
a
syntactico-semantic
aspect
of
linguistics
and
the
de
v
elopment
stages
of
machine
learning
for
NLP
.
The
study
suggests
that
syntactic-se
mantic
concepts
(semantic
role
labeling)
ha
v
e
been
le
v
er
-
aged
through
recent
trends
in
machine
learning
algorithms
and
may
benet
as
a
ne
w
paradigm
of
language-
independent
processing.
The
study
e
xplored
a
comprehens
i
v
e
w
ork
on
the
P
aninian
aspect
of
language
pro-
cessing
with
the
latest
trends
in
deep
learni
ng.
Ho
we
v
er
,
in-depth
studies
are
needed
to
get
linguistic
insights
especially
to
understand
speak
er
and
listener
communication.
Researchers
who
w
ant
to
utilize
NLP
for
v
arious
purposes
in
their
eld
can
understand
the
o
v
erall
technical
status
and
the
main
technologies
of
NLP
through
this
paper
.
Our
study
demonstrates
t
hat
Karaka
theory
retains
linguistic
insights,
which
are
more
resilient
than
other
semantic
methods.
The
in
v
estig
ation
opens
a
wide
scope
of
research
to
unfold
deeper
linguistic
aspects
with
feasible
w
ays
of
unfolding
cognition
of
Karaka
in
real-life
man-machine
interaction.
REFERENCES
[1]
N.
Sri
v
asta
v
a,
G.
Hinton,
A.
Krizhe
vsk
y
,
I.
Sutsk
e
v
er
,
and
R.
Salakhutdino
v
,
“Dropout:
a
simple
w
ay
to
pre
v
ent
neural
netw
orks
from
o
v
ertting,
”
J
ournal
of
Mac
hine
Learning
Resear
c
h
,
v
ol.
15,
no.
1,
pp.
1929–1958,
2014.
[2]
“SAS
data
and
AI
solutions,
”
F
ive
AI
tec
hnolo
gies
.
[Online].
A
v
ailable:
https://www
.sas.com/en
in/insights/articles/analytics/
v
e-
ai-technologies.html.
[3]
V
.
N.
Gudi
v
ada
and
K.
Arbabif
ard,
“Open-source
libraries,
application
frame
w
orks,
and
w
orko
w
systems
for
NLP
,
”
in
Compu-
tational
Analysis
and
Under
standing
of
Natur
al
Langua
g
es
:
Principles,
Methods
and
Applications
,
Series
Handbook
of
Statistics,
Else
vier
,
2018,
pp.
31–50.
[4]
“Natural
language
processing:
What
it
is?
and
ho
w
does
it
w
ork?”
[Online].
A
v
ailable:
https://monk
e
ylearn.com/natural-language-
processing.
[5]
M.
Raza,
“Importance
of
NLP
in
data
science,
”
Medium,
2021.
[Online].
A
v
ailable:
https://medium.com/theleanprogrammer/importance-of-nlp-in-data-science-3e2fd2424b2d.
A
surve
y
on
no
vel
appr
oac
h
to
semantic
computing
for
domain
specic
.
.
.
(Anjali
Bohr
a)
Evaluation Warning : The document was created with Spire.PDF for Python.
8
❒
ISSN:
2252-8776
[6]
A.
Shatna
wi,
G.
Al
-Bdour
,
R.
Al-Qurran,
and
M.
Al-A
yyoub,
“
A
comparati
v
e
study
of
open
source
deep
learning
frame
w
orks,
”
in
2018
9th
International
Confer
ence
on
Information
and
Communication
Systems,
ICICS
2018
,
2018,
v
ol.
2018-January
,
pp.
72–77,
doi:
10.1109/IA
CS.2018.8355444.
[7]
M
.
Y
ao,
“F
our
approaches
to
natural
language
processing
and
understanding,
”
T
opBots
,
2017.
[Online].
A
v
ailable:
https://www
.topbots.com/4-dif
ferent-approaches-natural-language-processing-understanding.
[8]
J
.
Y
i,
T
.
Nasuka
w
a,
R.
Bunescu,
and
W
.
Niblack,
“Sentiment
analyzer:
e
xtracting
sentiments
about
a
gi
v
en
topic
using
natural
language
processing
techniques,
”
in
Pr
oceedings
-
IEEE
International
Confer
ence
on
Data
Mining
,
ICDM
,
2003,
pp.
427–434,
doi:
10.1109/icdm.2003.1250948.
[9]
N.
T
apaswi
and
S.
Jain,
“T
reebank
based
deep
grammar
acquisition
and
part-of-speech
tagging
for
Sanskrit
sentences,
”
in
2012
CSI
6th
International
Confer
ence
on
Softwar
e
Engineering
,
CONSEG
2012
,
2012,
pp.
1–4,
doi:
10.1109/CONSEG.2012.6349476.
[10]
F
.
Sha
and
F
.
Pereira,
“Shallo
w
parsing
with
conditional
random
elds,
”
in
Pr
oceedings
of
the
2003
Human
Langua
g
e
T
ec
hnolo
gy
Confer
ence
of
the
North
American
Chapter
of
the
Association
for
Computational
Linguistics,
HL
T
-N
AA
CL
2003
,
2003,
pp.
213–220,
doi:
10.3115/1073445.1073473.
[11]
R.
McDonald,
K.
Crammer
,
and
F
.
Pereira,
“Fle
xible
te
xt
se
gmentation
with
structured
multilabel
classication,
”
in
HL
T/EMNLP
2005
-
Human
Langua
g
e
T
ec
hnolo
gy
Confer
ence
and
Confer
ence
on
Empirical
Methods
in
Natur
al
Langua
g
e
Pr
ocessing
,
Pr
oceed-
ings
of
the
Confer
ence
,
2005,
pp.
987–994,
doi:
10.3115/1220575.1220699.
[12]
A.
Ritter
,
C.
Sam,
Mausam,
and
O.
Etzioni,
“Named
entity
recognition
in
tweets:
an
e
xperimental
study
,
”
in
EMNLP
2011
-
Confer
ence
on
Empirical
Methods
in
Natur
al
Langua
g
e
Pr
ocessing
,
Pr
oceedings
of
the
Confer
ence
,
2011,
pp.
1524–1534.
[13]
S.
K.
Mishra
and
G.
N.
Jha,
“Sanskrit
K
¯
araka
analyzer
for
machine
translation,
”
in
SPLASH
pr
oceeding
of
iSTRANS
,
2007,
pp.
224–225.
[14]
A.
Ilyas,
K.
Shahzad,
and
M.
Kamran
Malik,
“Emotion
detection
in
code-mix
ed
Roman
Urdu-English
te
xt,
”
A
CM
T
r
ansactions
on
Asian
and
Low-Resour
ce
Langua
g
e
Information
Pr
ocessing
,
v
ol.
22,
no.
2,
pp.
1–28,
2023.
[15]
P
.
Martha,
“The
proposition
bank:
An
annotated
corpus
of
semantic
roles,
”
Computational
Linguistics
,
v
ol.
31,
no.
1,
pp.
71–105,
2005.
[16]
E.
Benson,
A.
Haghighi,
and
R.
Barzilay
,
“Ev
ent
disco
v
ery
in
social
media
feeds,
”
in
A
CL-HL
T
2011
-
Pr
oceedings
of
the
49th
Annual
Meeting
of
the
Association
for
Computational
Linguistics:
Human
Langua
g
e
T
ec
hnolo
gies
,
2011,
v
ol.
1,
pp.
389–398.
[17]
A.
Bohra
and
N.
C.
Barw
ar
,
“Performance
e
v
aluation
of
w
ord
representation
techniques
using
deep
learning
methods,
”
in
Pr
o-
ceedings
of
the
2020
International
Confer
ence
on
Computing
,
Communic
ation
and
Security
,
ICCCS
2020
,
2020,
pp.
1–7,
doi:
10.1109/ICCCS49678.2020.9277190.
[18]
A.
Bharati,
R.
Sang
al,
and
V
.
Chaitan
ya,
“Natural
language
processing,
comple
xity
theory
and
logic,
”
in
F
oundations
of
Softwar
e
T
ec
hnolo
gy
and
Theor
etical
Computer
Science:
T
enth
Confer
ence
,
Bangalor
e
,
India
December
17–19,
1990
Pr
oceedings
10
,
1990,
v
ol.
472
LNCS,
pp.
410–420,
doi:
10.1007/3-540-53487-3
60.
[19]
L.
W
ang,
H.
W
ang,
and
S.
Y
u,
“Semantic
computing
and
language
kno
wledge
bases,
”
in
IOP
Confer
ence
Series:
Materials
Science
and
Engineering
,
2017,
v
ol.
242,
no.
1,
p.
12109,
doi:
10.1088/1757-899X/242/1/012109.
[20]
P
.
Sheu,
“Semantic
computing
and
cogniti
v
e
computing/informatics,
”
in
2017
IEEE
16th
International
Confer
ence
on
Co
gnitive
Informatics
&
Co
gnitive
Computing
(ICCI*
CC)
,
2017,
pp.
4–4,
doi:
10.1109/icci-cc.2017.8109801.
[21]
P
.
C
.-Y
.
S
heu,
Semantic
computing
.
H.
Y
u,
C.
V
.
Ramamoorth
y
,
A.
K.
Joshi,
L.
A.
Zadeh,
Ed.,
W
ile
y
Online
Library
,
2010.
[22]
S.-U.
P
ark,
“
Analysis
of
the
status
of
natural
language
processing
technology
based
on
deep
learning,
”
The
J
ournal
of
Bigdata
,
v
ol.
6,
no.
1,
pp.
63–81,
2021.
[23]
S.
T
ekir
and
Y
.
Bastanlar
,
“Deep
le
arning:
e
x
emplar
studies
in
natural
language
processing
and
computer
vision,
”
Data
Mining
-
Methods,
Applications
and
Systems
,
p.
1,
2020,
doi:
10.5772/intechopen.91813.
[24]
J
.
Bro
wniee,
“Deep
learning
for
natural
language
processing,
”
in
Mac
hine
Learning
Mystery
,
2017.
[25]
N.
Ag
arw
al,
“The
ultimate
guide
to
dif
ferent
w
ord
embedding
techniques
in
NLP
,
”
KDnug
g
ets
,
2022.
https://www
.kdnuggets.com/2021/11/guide-w
ord-embedding-techniques-nlp.html.
[26]
D
.
Suhartono,
A.
P
.
Gema,
S.
W
inton,
T
.
Da
vid,
M.
I.
F
anan
y
,
and
A.
M.
Arymurth
y
,
“
Ar
gument
annotation
and
analysis
using
deep
learning
with
attention
mechanism
in
Bahasa
Indonesia,
”
J
ournal
of
Big
Data
,
v
ol.
7,
no.
1,
pp.
1–18,
2020,
doi:
10.1186/s40537-
020-00364-z.
[27]
T
.
Katte,
“Recurrent
neural
netw
ork
and
its
v
arious
architecture
types,
”
International
J
ournal
of
Resear
c
h
and
Scientic
Inno
vation
(IJRSI)
,
v
ol.
5,
no.
3,
pp.
124–129,
2018.
[28]
R.
Salakhutdino
v
and
G.
Hinton,
“Deep
Boltzmann
machines,
”
in
Pr
oceedings
of
the
T
welfth
International
Confer
ence
on
Articial
Intellig
ence
and
Statistics
,
2009,
v
ol.
5,
pp.
448–455.
[29]
N.
Sethi,
A.
De
v
,
and
P
.
Bansal,
“
A
no
v
el
neural
machine
translation
approach
for
lo
w-resource
Sanskrit-Hindi
language
pair
,
”
A
CM
T
r
ansactions
on
Asian
and
Low-Resour
ce
Langua
g
e
Information
Pr
ocessing
,
Apr
.
2023,
doi:
10.1145/3591207.
[30]
L.
Ferrone
and
F
.
M.
Zanzotto,
“Symbolic,
distrib
uted,
and
distrib
utional
representations
for
natural
language
processing
in
the
era
of
deep
learning:
a
surv
e
y
,
”
F
r
ontier
s
in
Robotics
and
AI
,
v
ol.
6,
p.
153,
2020,
doi:
10.3389/frobt.2019.00153.
[31]
H.
Li,
“Deep
learning
for
natural
language
processing:
adv
antages
and
challenges,
”
National
Science
Re
vie
w
,
v
ol.
5,
no.
1,
pp.
24–26,
2018,
doi:
10.1093/nsr/nwx110.
[32]
A.
Bharati,
V
.
Chaitan
ya,
R.
Sang
al,
and
B.
Gillon,
Natur
al
langua
g
e
pr
ocessing:
A
P
aninian
per
spective
.
Prentice
Hal
l
India
Learning
Pri
v
ate
Limited,
1995.
[33]
S.
Selot,
N.
T
ripathi,
and
A.
S.
Zadg
aonkar
,
“Neural
netw
ork
model
for
semantic
analysis
of
Sanskrit
te
xt,
”
International
J
ournal
of
Natur
al
Computing
Resear
c
h
,
v
ol.
7,
no.
1,
pp.
1–14,
2018,
doi:
10.4018/ijncr
.2018010101.
[34]
S.
D.
Joshi,
“Syntactic
and
semantic
de
vices
in
the
Ast
¯
adh
y
¯
ay
¯
ı
of
P
¯
an
.
ini,
”
J
ournal
of
Indian
Philosophy
,
v
ol.
29,
no.
1–2,
pp.
155–167,
2001,
doi:
10.1023/a:1017559527007.
[35]
S.
C.
Kak,
“The
P
aninian
approach
to
natural
language
processing,
”
International
J
ournal
of
Appr
oximate
Reas
oning
,
v
ol.
1,
no.
1,
pp.
117–130,
1987,
doi:
10.1016/0888-613X(87)90007-7.
[36]
P
.
R.
Chaudhari,
P
.
C.
Gangurde,
and
N.
L.
K
ulkarni,
“Study
of
methodologies
for
util
izing
Sanskrit
in
computational
linguistics,
”
International
J
ournal
of
Electr
onics,
Communication
and
Soft
Computing
Science
&
Engineering
,
p.
1,
2015.
Int
J
Inf
&
Commun
T
echnol,
V
ol.
14,
No.
1,
April
2025:
1–10
Evaluation Warning : The document was created with Spire.PDF for Python.
Int
J
Inf
&
Commun
T
echnol
ISSN:
2252-8776
❒
9
[37]
D.
Jha,
R.
Jha,
and
V
.
V
arshne
y
,
“Na
tural
language
processing
and
Sanskrit,
”
International
J
ournal
of
Computer
Engineering
&
T
ec
hnolo
gy
,
v
ol.
5,
no.
10,
pp.
57–63,
2014.
[38]
Amita
and
A.
Jangra,
“
An
annotation
scheme
for
English
language
using
P
ani
nian
frame
w
ork,
”
IJISET
-
International
J
ournal
of
Inno
vative
Science
,
Engineering
&
T
ec
hnolo
gy
,
v
ol.
2,
no.
1,
pp.
616–619,
2015.
[39]
A.
Bharati
and
R.
Sang
al,
“Computational
P
aninian
grammar
frame
w
ork,
”
in
Superta
g
ging:
Using
comple
x
le
xical
descriptions
in
natur
al
langua
g
e
pr
ocessing
,
2010,
pp.
355–370.
[40]
S.
Chatterji,
T
.
M.
Sarkar
,
S.
Sarkar
,
and
J.
Chakraborty
,
“Karak
relations
in
Beng
ali,
”
in
Pr
oceedings
of
31st
All-India
Confer
ence
of
Linguists
(AICL
2009)
,
2009,
pp.
33–36.
[41]
B.
V
.
S.
Perraju,
“
Algorithmic
aspects
of
natural
language
parsing
using
P
aninian
frame
w
ork,
”
M.
T
ech
thesis,
Department
of
Computer
Science
and
Engineering,
IIT
Kanpur
,
1992.
[42]
K.
Alla,
R.
Prasad,
and
R.
R.
Kandula,
“
A
ne
w
approach
to
Hindi
te
xt
ste
g
anograph
y
using
Hindi
Karak
Kriyaye,
”
J
ournal
of
Information
Assur
ance
&
Security
,
v
ol.
6,
no.
6,
2011.
[43]
A.
Bada
wi,
“The
ef
fecti
v
eness
of
natural
language
processing
(NLP)
as
a
processing
solut
ion
and
semantic
impro
v
ement,
”
Interna-
tional
J
ournal
of
Economic,
T
ec
hnolo
gy
and
Social
Sciences
,
v
ol.
2,
no.
1,
pp.
36–44,
2021,
doi:
10.53695/injects.v2i1.194.
[44]
S.
Li,
J.
Zhang,
X.
Huang,
S.
Bai,
and
Q.
Liu,
“Semantic
computation
in
a
Chinese
question-answering
system,
”
J
ournal
of
Computer
Science
and
T
ec
hnolo
gy
,
v
ol.
17,
no.
6,
pp.
933–939,
2002,
doi:
10.1007/BF02960786.
[45]
S.
A.
Salloum,
R.
Khan,
and
K.
Shaalan,
“
A
surv
e
y
of
semantic
analysis
approaches,
”
in
Pr
oceedings
of
the
International
Confer
ence
on
Articial
Intellig
ence
and
Computer
V
ision
(AICV2020)
,
2020,
v
ol.
1153
AISC,
pp.
61–70,
doi:
10.1007/978-3-030-44289-7
6.
[46]
T
.
Landauer
and
S.
Dumais,
“Latent
semantic
analysis,
”
Sc
holarpedia
,
v
ol.
3,
no.
11,
p.
4356,
2008,
doi:
10.4249/scholarpedia.4356.
A
v
ailable:
http://www
.scholarpedia.or
g/article/Latentsemanticanalysis?ref=https://githubhelp.com
[47]
A.
Bharati,
M.
Bhatia,
V
.
Chaitan
ya,
and
R.
Sang
al,
“P
aninian
grammar
frame
w
ork
applied
to
English,
”
Department
of
Computer
Science
and
Engineering,
Indian
Institute
of
T
echnology
,
Kanpur
,
1996.
[48]
S.
Mittal,
“
A
re
vie
w
of
Engli
sh
to
Indian
language
translator:
Anusaaraka,
”
IJCA
Pr
oceedings
on
International
Confer
ence
on
Advances
in
Computer
Engineering
and
Applications
,
v
ol.
ICA
CEA,
no.
1,
pp.
1–6,
2014.
[49]
A.
P
.
K
ulkarni,
“Design
and
architecture
of
‘
Anusaaraka’-An
approach
to
machine
translation,
”
Satyam
T
ec
hnical
Re
vie
w
,
v
ol.
3,
pp.
57–64,
2003.
[50]
S.
P
aul,
M.
T
andon,
N.
Joshi,
and
I.
Mathur
,
“Design
of
a
rule
based
Hindi
lemmatizer
,
”
in
Pr
oceedings
of
Thir
d
Interna-
tional
W
orkshop
on
Articial
Intellig
ence
,
Soft
Computing
and
Appli
cations,
Chennai,
India
,
2013,
v
ol.
2,
pp.
67–74,
doi:
10.5121/csit.2013.3408.
[51]
P
.
Katyayan
and
N.
Jos
hi,
“De
v
elopment
of
automatic
rule-based
semantic
tagger
and
Karaka
analyzer
for
Hindi,
”
T
r
ansactions
on
Asian
and
Low-Resour
ce
Langua
g
e
Information
Pr
ocessing
,
v
ol.
21,
no.
2,
pp.
1–25,
2021,
doi:
10.1145/3479155.
[52]
A.
Das,
T
.
Halder
,
and
D.
Saha,
“
Automatic
e
xtraction
of
Beng
ali
root
v
erbs
using
P
aninian
grammar
,
”
in
RTEICT
2017
-
2nd
IEEE
International
Confer
ence
on
Recent
T
r
ends
in
Electr
onics,
Information
and
Communication
T
ec
hnolo
gy
,
Pr
oceedings
,
2017,
v
ol.
2018-January
,
pp.
953–956,
doi:
10.1109/R
TEICT
.2017.8256739.
[53]
B.
K.
Mishra
and
R.
Sahoo,
“
A
h
ybrid
kno
wledge
mining
approach
to
de
v
elop
a
system
frame
w
ork
for
Odia
language
te
xt
process-
ing,
”
Materials
T
oday:
Pr
oceedings
,
v
ol.
5,
no.
1,
pp.
1335–1340,
2018,
doi:
10.1016/j.matpr
.2017.11.219.
[54]
J.
Bronkhors
t,
“The
role
of
meanings
in
P
anini’
s
grammar
,
”
Indian
Linguistics
,
v
ol.
40,
pp.
146–157,
1979.
[55]
S.
Bhate
and
S.
Kak,
“P
¯
an
.
ini’
s
grammar
and
computer
science,
”
Annals
of
the
Bhandarkar
Oriental
Resear
c
h
Institute
,
v
ol.
72,
no.
1/4,
pp.
79–94,
1991.
[56]
P
.
Lahoti,
N.
Mittal,
and
G.
Singh,
“
A
surv
e
y
on
NLP
resources,
tools,
and
techniques
for
Marat
hi
language
processing,
”
A
CM
T
r
ansactions
on
Asian
and
Low-Resour
ce
Langua
g
e
Information
Pr
ocessing
,
v
ol.
22,
no.
2,
pp.
1–34,
2022,
doi:
10.1145/3548457.
[57]
A.
Kataria
and
R.
Nath,
“Natural
language
interf
ace
for
databases
in
Hindi
based
on
Karaka
theory
,
”
International
J
ournal
of
Computer
Applications
,
v
ol.
122,
no.
7,
pp.
39–43,
2015,
doi:
10.5120/21716-4841.
[58]
A.
Y
ajnik
and
D.
Sharma,
“P
arsi
ng
techniques
using
P
aninian
frame
w
ork
on
Nepali
language,
”
DJ
J
ournal
of
Engineering
and
Applied
Mathematics
,
v
ol.
1,
no.
1,
pp.
23–28,
2015,
doi:
10.18831/djmaths.or
g/2015011004.
[59]
N.
Kanuparthi,
A.
Inumella,
and
D.
M.
Sharma,
“Hindi
deri
v
ational
morphological
analyzer
,
”
in
Pr
oceedings
of
the
T
welfth
Meeting
of
the
Special
Inter
est
Gr
oup
on
Computational
Morpholo
gy
and
Phonolo
gy
,
pp.
10–16,
2012.
[60]
A.
Das,
“Opinion
e
xtraction
and
summarization
from
te
xt
documents
in
Beng
ali,
”
Department
of
Compute
r
Science
&
Engineering,
Jada
vpur
Uni
v
ersity
,
2011.
[61]
V
.
N.
Mi
sra,
The
descriptive
tec
hnique
of
P
an
.
ini:
an
intr
oduction
,
v
ol.
18.
W
alter
de
Gruyter
GmbH
&
Co
KG,
2017.
[62]
S.
Muzaf
f
ar
,
P
.
Behera,
and
G.
N.
Jha,
“
A
P
¯
anini
¯
an
frame
w
ork
for
analyzing
case
mark
er
errors
in
English-Urdu
machine
translation,
”
Pr
ocedia
Computer
Science
,
v
ol.
96,
pp.
502–510,
2016,
doi:
10.1016/j.procs.2016.08.117.
S.
K.
Mishra
and
G.
N.
Jha,
“Sanskrit
K
¯
araka
analyzer
for
machine
translation,
”
in
SPLASH
pr
oce
eding
of
iSTRANS
,
2007,
pp.
224–225.
[63]
L.
Sindhu,
“
An
inte
grated
approach
for
plagiarism
detection
in
Malayalam
documents,
”
Cochin
Uni
v
ersity
of
Science
and
T
echnol-
ogy
,
2017.
[64]
H.
Redkar
et
al.
,
“V
erbframator:
Semi-automatic
v
erb
frame
annotator
tool
with
special
reference
to
Marathi,
”
in
Pr
oceedings
of
the
13th
International
Confer
ence
on
Natur
al
Langua
g
e
Pr
ocessing
,
2016,
pp.
299–304.
[65]
K.
Anuranjana,
V
.
A.
Rao,
and
R.
Mamidi,
“Hindi
question
gene
ration
using
dependenc
y
structures,
”
arXi
v
preprint
arXi
v:1906.08570,
2019,
[Online].
A
v
ailable:
http://arxi
v
.or
g/abs/1906.08570.
[66]
S.
M.
Archana,
N.
V
ahab,
R.
Thankappan,
and
C.
Raseek,
“
A
rule
based
quest
ion
answering
system
in
Malayalam
corpus
using
V
ibhakthi
and
POS
tag
analysis,
”
Pr
ocedia
T
ec
hnolo
gy
,
v
ol.
24,
pp.
1534–1541,
2016,
doi:
10.1016/j.protc
y
.2016.05.124.
P
.
Katyayan
and
N.
Joshi,
“De
v
elopment
of
automatic
rule-based
semantic
tagger
and
Karaka
analyzer
for
Hindi,
”
T
r
ansactions
on
Asian
and
Low-Resour
ce
Langua
g
e
Information
Pr
ocessing
,
v
ol.
21,
no.
2,
pp.
1–25,
2021,
doi:
10.1145/3479155.
[67]
S.
Sharma
and
V
.
Gupta,
“Punjabi
documents
clustering
system,
”
J
ournal
of
Emer
ging
T
ec
hnolo
gie
s
in
W
eb
Intellig
ence
,
v
ol.
5,
no.
2,
pp.
171–187,
2013,
doi:
10.4304/jetwi.5.2.171-187.
[68]
J.
P
.
Jayan,
J.
S.
K
umar
,
and
T
.
Amudha,
“Semantic
role
identication
for
Malayalam
using
machi
ne
learning
approaches,
”
Inno
va-
tions
in
Systems
and
Softwar
e
Engineering
,
pp.
1–7,
2023,
doi:
10.1007/s11334-022-00496-w
.
A
surve
y
on
no
vel
appr
oac
h
to
semantic
computing
for
domain
specic
.
.
.
(Anjali
Bohr
a)
Evaluation Warning : The document was created with Spire.PDF for Python.
10
❒
ISSN:
2252-8776
[69]
K.
T
.
Radhika
and
P
.
C.
R.
Raj,
“Semantic
role
e
xtraction
and
general
concept
understanding
in
Malayalam
using
P
aninian
grammar
,
”
International
J
ournal
of
Engineering
Resear
c
h
and
De
velopment
,
v
ol.
9,
no.
3,
pp.
28–33,
2013.
[70]
D.
Rai,
“KarakA
CR
OSS:
a
Karaka-based
approach
to
cross-lingual
sentiment
analysis,
”
International
J
ournal
of
Liter
atur
e
and
Langua
g
es
,
v
ol.
3,
no.
09,
pp.
1–4,
2023,
doi:
10.37547/ijll/V
olume03Issue09-01
01-04
0.
[71]
M
.
Rahul
Raj,
R.
P
.
Haroon,
and
N.
V
.
Sobhana,
“
A
no
v
el
e
xtracti
v
e
te
xt
summarization
system
with
self-or
g
anizing
map
clustering
and
entity
recognition,
”
Sadhana
-
Academy
Pr
oceedings
in
Engineering
Sciences
,
v
ol.
45,
no.
1,
p.
32,
2020,
doi:
10.1007/s12046-
019-1248-0.
[72]
J
.
A.
Gajjam,
“T
aking
a
road
not
tra
v
elled:
A
cross-lingual
study
of
K
¯
araka-cognition,
”
Langua
g
e
in
India
,
v
ol.
21,
no.
5,
pp.
1–24,
2021.
[73]
B
.
Lahiri,
The
case
system
of
Eastern
Indo-Aryan
langua
g
es:
a
typolo
gical
o
vervie
w
.
Routledge
India,
2021.
[74]
D.
A.
V
erma,
R.
S.
Joshi,
S.
A.
Joshi,
and
O.
K.
Susladkar
,
“Study
of
similarity
measures
as
features
in
classication
for
answer
sentence
selection
task
in
Hindi
question
answering:
language-specic
vs
other
measures,
”
in
Pr
oceedings
of
the
35th
P
acic
Asia
Confer
ence
on
Langua
g
e
,
Information
and
Computation,
P
A
CLIC
2021
,
2021,
pp.
715–724.
[75]
J
.
P
.
Jayan
and
V
.
Go
vindaru,
“
Automatic
summarization
of
Malayalam
documents
using
te
xt
e
xtraction
methods,
”
in
SCRS
Confer
-
ence
Pr
oceedings
on
Intellig
ent
Systems
,
2021,
pp.
443–457,
doi:
10.52458/978-93-91842-08-6-42.
[76]
D.
V
erma,
R.
Joshi,
A.
Shi
v
ani,
and
R.
Gupta,
“Karaka-based
answer
retrie
v
al
for
question
answering
in
Indic
languages,
”
in
International
Confer
ence
Recent
Advances
in
Natur
al
Langua
g
e
Pr
ocessing
,
RANLP
,
2023,
pp.
1216–1224,
doi:
10.26615/978-
954-452-092-2
129.
[77]
A.
V
aidya,
S.
Husain,
P
.
Mannem,
and
D.
M.
Sharma,
“
A
Karaka
based
annotation
scheme
for
English,
”
in
Computational
Li
nguistics
and
Intellig
ent
T
e
xt
Pr
ocessing:
10th
International
Confer
ence
,
CICLing
2009,
Me
xico
City
,
Me
xico,
Mar
c
h
1-7,
2009.
Pr
oceedings
10
,
2009,
v
ol.
5449
LNCS,
pp.
41–52,
doi:
10.1007/978-3-642-00382-0
4.
[78]
A.
Bharati
and
R.
Sang
al,
“P
arsing
free
w
ord
order
languages
in
the
P
aninian
frame
w
ork,
”
in
31st
Annual
Meeting
of
the
Association
for
Computational
Linguistics
,
1993,
v
ol.
1993-June,
pp.
105–111,
doi:
10.3115/981574.981589.
[79]
R.
Briggs,
“Kno
wledge
representation
in
Sanskrit
and
articial
intelligence,
”
AI
Ma
gazine
,
v
ol.
6,
no.
1,
p.
32,
1985,
[Onl
ine].
A
v
ailable:
http://www
.aaai.or
g/ojs/inde
x.php/aimag
azine/article/vie
w/466/402.
[80]
S.
Singh
and
T
.
J.
Siddiqui,
“Role
of
Karaka
relations
in
Hindi
w
ord
sense
disambiguation,
”
J
ournal
of
Information
T
ec
hnolo
gy
Resear
c
h
(JITR)
,
v
ol.
8,
no.
3,
pp.
21–42,
2015,
doi:
10.4018/JITR.2015070102.
[81]
R
.
Zhao,
R.
Y
an,
Z.
Chen,
K.
Mao,
P
.
W
ang,
and
R.
X.
Gao,
“Deep
learning
and
its
applications
to
machine
health
monitoring,
”
Mec
hanical
Systems
and
Signal
Pr
ocessing
,
v
ol.
115,
pp.
213–237,
2019,
doi:
10.1016/j.ymssp.2018.05.050.
[82]
S
.
K.
Karn,
“Case
system
in
Maithili:
from
semantic
perspecti
v
e,
”
Nepalese
Linguistics
,
v
ol.
23,
p.
126,
2008.
[83]
“
Natural
language
processing
tutorial,
”
tutorialspoint,
[Online].
A
v
ailable:
https://www
.tutorialspoint.com/natural
language
processing/inde
x.htm.
[84]
O.
Rambo
w
,
B.
Dorr
,
K.
Kipper
,
I.
K
u
ˇ
cero
v
´
a,
and
M.
P
almer
,
“
Automa
tically
deri
ving
tectogrammatical
labels
from
other
resources:
a
comparison
of
semantic
labels
across
frame
w
orks,
”
Pr
a
gue
Bulletin
of
Mathematical
Linguistics
,
v
ol.
79–80,
pp.
23–35,
2003.
[85]
P
.
J.
Anton
y
and
K.
P
.
Soman,
“Computational
morphology
and
natural
language
parsing
for
Indian
languages:
a
literature
surv
e
y
,
”
International
J
ournal
of
Scientic
&
Engineering
Resear
c
h
,
v
ol.
3,
no.
3,
pp.
136–146,
2012.
[86]
A
.
Bharati
and
A.
K
ulkarni,
“Information
coding
in
a
language:
some
insights
from
P
aninian
grammar
,
”
2008.
[87]
K
.
P
arton
et
al.
,
“Who,
what
,
when,
where,
wh
y?
Comparing
multiple
approaches
to
the
cross-lingual
5W
task,
”
in
A
CL-IJCNLP
2009
-
J
oint
Confer
ence
of
the
47th
Annual
Meeting
of
the
Association
for
Computational
Linguistics
and
4th
International
J
oint
Confer
ence
on
Natur
al
Langua
g
e
Pr
ocessing
of
the
AFNLP
,
Pr
oceedings
of
the
Confer
ence
,
2009,
pp.
423–431.
[88]
A.
Bharati
and
R.
Sang
al,
“
A
Karaka
based
approach
to
parsing
of
Indian
languages,
”
in
COLING
1990
V
olume
3:
P
aper
s
pr
esented
to
the
13th
International
Confer
ence
on
Computational
Linguistics
,
1990,
pp.
25–29,
doi:
10.3115/991146.991151.
BIOGRAPHIES
OF
A
UTHORS
Anjali
Bohra
recei
v
ed
her
rst
de
gree
Bachelor
of
Engineering
in
Computer
Science
and
Engineering
from
Mody
Colle
ge
of
Engineering
and
T
echnology
,
Rajasthan
Uni
v
ersity
,
Laxmang
arh,
Rajasthan
2002.
She
has
also
attained
Master
de
gree
in
Computer
Science
and
Engineering
from
MBM
Engineering
Colle
ge,
Jai
Narain
Vyas
Uni
v
ersity
,
Jodhpur
,
Rajasthan
i
n
2012.
Currently
a
Ph.D.
scholar
and
her
research
interests
focus
on
natural
language
processing,
articial
intelligence,
machine
learning,
and
deep
learning.
She
can
be
contacted
at
email:
anjali
vb
.phdcse@mbm.ac.in.
Nemi
Chand
Barwar
has
B.E.
in
Computer
T
echnology
from
MANIT
Bhopal,
M.E.
in
Digital
Communication,
and
a
Ph.D.
from
MBM
Engineering
Colle
ge,
Jodhpur
.
He
w
orks
as
a
Professor
at,
the
Department
of
Computer
Science
&
Engineering,
MBM
Uni
v
ersity
,
Jodhpur
.
He
has
e
xperience
of
o
v
er
30
years
in
the
eld
of
teaching
and
research.
He
has
published
more
than
60
research
papers
in
national
and
international
conferences
a
nd
journals.
He
is
supervising
the
Ph.D.
research
program
in
computer
science
and
engine
ering
discipline
as
well
as
in
Information.
His
research
and
teaching
interests
are
computer
netw
orking,
WSN,
MANET/V
ANET
,
IoT
,
big
data
analytics,
V
oD,
P2P
netw
orks
,
and
machine
learning.
He
had
or
g
a
nized
10
national
conferences
and
short-term
courses
sponsored
by
AICTE/UGC/DST
.
He
is
a
l
ife
member
of
ISTE,
IEI.
He
can
be
contacted
at
email:
nemi.cse@mbm.ac.in.
Int
J
Inf
&
Commun
T
echnol,
V
ol.
14,
No.
1,
April
2025:
1–10
Evaluation Warning : The document was created with Spire.PDF for Python.