Jurnal RESTI (Rekayasa Sistem dan Teknologi Informasi),
Год журнала:
2023,
Номер
7(4), С. 817 - 823
Опубликована: Авг. 12, 2023
On
Twitter,
users
can
post
tweets,
videos,
and
images.
It
can,
however,
also
be
disruptive
difficult.
To
categorize
the
material
improve
searchability,
hashtags
are
crucial.
This
study
focuses
on
examining
opinions
of
Twitter
who
participate
in
trending
topics.
The
algorithms
K-Nearest
Neighbor
(KNN)
Support
Vector
Machine
(SVM)
used
for
sentiment
analysis.
data
set
comprises
tweet
information
popular
topics
that
was
collected
using
API
saved
Excel
format.
SVM
K-NN
preparation,
weighting,
With
105
points,
provides
insight
into
user
sentiment.
identified
99%
positive
responses
1%
negative
with
an
accuracy
80%.
KNN
successfully
90%
10%
responses,
rate
71.4%.
According
to
results,
performs
better
when
analyzing
hashtag
Twitter.
Natural Language Processing Journal,
Год журнала:
2024,
Номер
6, С. 100059 - 100059
Опубликована: Фев. 29, 2024
Sentiment
analysis
is
a
method
within
natural
language
processing
that
evaluates
and
identifies
the
emotional
tone
or
mood
conveyed
in
textual
data.
Scrutinizing
words
phrases
categorizes
them
into
positive,
negative,
neutral
sentiments.
The
significance
of
sentiment
lies
its
capacity
to
derive
valuable
insights
from
extensive
data,
empowering
businesses
grasp
customer
sentiments,
make
informed
choices,
enhance
their
offerings.
For
further
advancement
analysis,
gaining
deep
understanding
algorithms,
applications,
current
performance,
challenges
imperative.
Therefore,
this
survey,
we
began
exploring
vast
array
application
domains
for
scrutinizing
context
existing
research.
We
then
delved
prevalent
pre-processing
techniques,
datasets,
evaluation
metrics
comprehension.
also
explored
Machine
Learning,
Deep
Large
Language
Models
Pre-trained
models
providing
advantages
drawbacks.
Subsequently,
precisely
reviewed
experimental
results
limitations
recent
state-of-the-art
articles.
Finally,
discussed
diverse
encountered
proposed
future
research
directions
mitigate
these
concerns.
This
review
provides
complete
covering
models,
domains,
challenges,
directions.
Intelligent Systems with Applications,
Год журнала:
2023,
Номер
19, С. 200256 - 200256
Опубликована: Июль 13, 2023
The
existing
multimodal
biometric
fingerprint
and
vein
deep
learning
features
were
found
effective
for
recognition.
However,
the
current
performance
of
was
limited
due
to
missing
temporal
image
dependence
extracted
can
obfuscate
some
important
information
because
irrelevant
features.
This
work
proposes
a
sequence
filtered
spatial
finger
veins
network
(FS-STMFPFV-Net).
overall
proposed
FS-STMFPFV-Net
is
achieved
from
two-channel
independent
improve
variabilities.
In
first
channel,
generated
by
aligning
images
together
inside
generator.
sequences
are
built
into
five
layers
convolution
neural
fusion
model
extract
sequence-wise
second
channel
where
remembered
in
long
short-term
memory
interactions
between
dimensions
which
finally
generate
their
long-term
dependencies
as
complementary
information.
These
fused
together,
discriminative
selected
using
feature
selection.
We
have
presented
ReliefFS
selection
serve
basis
selecting
compact
CNN-based
To
evaluate
FS-STMFPFV-Net,
NUPT-FPV,
FVC-2002-DBs,
CASIA
dataset
used,
provides
fingerprint,
finger-vein,
palmprint
databases,
experimental
validation.
when
evaluated
standard
protocols
offers
more
than
97%
accuracy
across
different
databases
ten
times
computationally
friendly
algorithms.
This
paper
presents
a
study
on
the
use
of
Chat
Generative
Pretrained
Transformer
(ChatGPT)
in
education.
In
this
work,
we
propose
sentiment
analysis
model
tweets
related
to
ChatGPT
The
purpose
research
is
identify
common
sentiments,
topics,
and
perspectives
that
are
expressed
towards
education
field
based
data
collected
from
Twitter.
Twitter
was
used
collect
11830
about
Topics
emotions
were
extracted
using
NLP
algorithms
organized
into
distinct
groups.
Also,
most
frequent
words
positive
negative
opinion
determined.
findings
indicate
either
or
neutral,
with
small
percentage
expressing
sentiments.
addition,
analyzes
sentiments
employment
four
different
classifiers:
Naive
Bayes
(NB),
Support
Vector
Machine
(SVM),
K-Nearest
Neighbors
(KNN),
Random
Forest
(RF).
According
results,
SVM
classifier
has
highest
accuracy
81.4
percent.
2022 International Conference on Inventive Computation Technologies (ICICT),
Год журнала:
2024,
Номер
unknown
Опубликована: Апрель 24, 2024
Sentiment
analysis
has
emerged
as
a
pivotal
tool
for
distilling
valuable
insights
from
the
vast
expanse
of
textual
information,
significantly
influencing
financial
markets.
Traditional
models,
however,
often
fall
short
in
navigating
complex
and
nuanced
terrain
economic
texts,
struggling
to
decode
industry-specific
jargon
rapid
shifts
sentiment
inherent
news.
This
discrepancy
highlights
notable
gap
applying
general
linguistic
models
domain-specific
scenarios,
presenting
challenges
leveraging
market
prediction
strategic
decision-making.
To
bridge
this
divide,
refined
approach
is
introduced,
featuring
an
advanced
adaptation
Llama
2
7b-hf
model.
model
specifically
fine-tuned
domain,
employing
parameter-efficient
fine-tuning
(PEFT)
methods
Simple
Fine-tuning
Trainer
(SFTTrainer).
Such
modifications
enhance
model's
attunement
lexicon
subtleties,
ensuring
precise
sector's
distinct
characteristics
while
avoiding
catastrophic
forgetting.
Demonstrated
results
signify
considerable
enhancements
accuracy
within
sector.
The
achieves
overall
89%,
marking
substantial
improvements
across
negative,
neutral,
positive
categories
when
juxtaposed
with
baseline
counterparts.
Accuracy
elevates
37.3%
conditions
84.4%
post-initial
adjustments,
culminating
at
89%
after
comprehensive
fine-tuning,
affirming
enhanced
proficiency
decoding
dynamics
sentiment.
