unit -c

Aclvance d sentiment onalysis

Inbodvetion to advanced sentimenb enalysis :

classity arias samples of

*Sentiment analysis s the practice of using algoribhrns to
velated bexb into overall posibive and negative categories
can help pu determne the ratio of posibive to ncgative egagemants
"sentiment analysis
about a specific topic .
follouing comments, twects, and product
" n this presentation,we will explore the
veviews, and deep leavning.
: ocume nt-level ainrs to classty the entire text as positive, neqotive,
" Oocunént Level
O nevtal
paybs.
:Understanding of bhe sentbiment expressed in diferent tert
" Sentence level entibies or targets
level:zdenbifies the sentiment erpressed towards specific
" Entity- cornpanies.
mentioned in the text,such as people , the sentiment
sentiment level : this approach involves compating
"compavative
betueen differcnt entities.

Major challenges:
hunan larguoge and understonding is rich and intiicated
-Language diffevences :he
poken by humans
and there many languages
collecbion of input-output poirs,tere
.TYaining data : TYaining data is a curated
input Yepfe serntrs the eatures ¡ atbributes of the data.
the
task : Task such as cla 3stHcabion of text or analyzing the sentiment
comeleitu of the
kss time compared to moie cornplex tasks such as machie
of the text may require
tiarslation of answering the questions·
lack of scalabiliby
Limitations :piouty in honding armbtguity
 Small model in NLP :
that have beern trained on huge amaunts of
models
"frebiained modek ate deep leaning
data be fore fine- buving fo a specitic
task.
:(Bidinectional Encodet Representations from Thansfor mers)
"BERT based madel
tanstormer ) : GPT- i9 a tronsformer-
" GPT-2 : (Genevatie pretioined
selr-supervised manner
english corpus in a
Preb rained on an extensive transtormer mdel, wtbich includes Telate
XL is avatient of
"TYanstoimer -XL : TYansformer -
encoding and a Yecuivene mechanis m.
Posibional

Build a sentirnent analysis : implement .Its based on manually

the basic and easy appoach to
"Rule -based : this s
syster analyze the text it veads
the The sysberm is first
Pre-defned vules, helping both NPL and ML-
:this is the advanced appioach, using either negative,neutural,
" Automabic pre-defied as
thousands of expressiors that are
fed with
or posibive. effective algorithm. This approach
most
:zts the best of both worlds and the while vunning through new
terms
"Hybid algoribhn
accUracy of the ule-based
enjoys the high
the blink of an eye
and Cxpressions in
analsis isnt Just to customey.
:sentiment
" challerges
networks (CNNs):
Model: Con volutional neural
netwoks (ANN) which is pred orninantly
the extended version of artificial neural mabrix dataseb.
CNN iS ike
extiact the feature from the gid- convdlubioral laye,
USed to
multiple layers like, the inpub layer,
:CNN orsists of
"CNN avthitectuve connected layers.
pooling layer,and. lly t peutal netwotks CRNNS), Conolutional neural
nebwork aychitectures:
NeUYal CCNNs)and, transfoimers
" netwotks

and require a lot of merory.

comptationally expensive to train
"challenges
 bo as;ti-gram).
important
the vwiby
as
the rble such subjertione
setiment), thi impuses
neual meas
in is tasks love, hatehuge
leatning
sed (Bigram and rmore
sadness, a helpsand
ond that four to
architecture processing sentiment bit theres
interface"and
Hochrelter
(RNN) contans an negative
ave this machine motives
words
summarization.
is chssifiers method
network anger, that
for
thas language of present' the
(RNN)sepp gioupings of ond available advanced
weight fear, the
obvious
This ze
neural stucbure posibive
gained onalg
network
by cels.ral n-qram hate".
ke
natutext small the Yesources more like ib's
: evern
celk I49 YecuIteot don't
in
ncual chaincalledfor ond o1
Custfine
Understanding usesernotions sut and and
LsTy proposed o elassificatior
thatmang :"Inegative.
ike"
used emotion,
vecurrernt s has.ablocks lation words
clauses tecognition staterments
(RNN)rth been is cormplex dont
JBLSTM
both.architecture tr£ns of analyzedmyselfot
both
was memoty
have usingbag is,
analysis
sertimert
Fine-qveined ave an
arbificial
It in machine grain" binary emcion
more ae C dterence
hetween o
within
buork learningLsTNba :LSTNS the by there mo1e. two
being
schmdhuber.
Jurgen
BI da
different
LSTA sentencebekHind use faced analyze compase
they
product"
beface
ne an LSTM:
seqvential Modelin9modeling, also
detetion,
ov rmany i
onalysis thernselves
words tent which analysis:
Emotion
neuralis LsTA: as
jntuibion and subtleties
LSTMdcep "Bdirectionnl and the the Yeason
classifiers
challenging to and we fCusbomers.
RecuTrent lanquage
Language this in , :f
techniques
of
LSTN: of
Process networks oUnd fustation
Held woking whereThe good Emotiorn EXampe
zn hese
) a) a)
. "
 gather
opinion mining :
analysis, includes building a fame bork to utilzes
senbiment
)which is likewise called Computerized opinion mining vegularly
item
and artange opinions about anartifical intelligence CA),to dig content for
machine taking in, a kind of
ment tight
sentiment accomplishment of an advance
enable advertiseYs to assess the adjustments of a thing or organization
)It can
iten dispatch, nake sense of which
ot new
ate well known.

sentiment analysis technique : (NEP) bßchaiquss

RHbutoTSAtoato tosqu6gespiocssstr Rule-based approach
) Lericon-based qpproach sentiment analysis (ABSA)
app1oach s) Aspect-based
learring-based c) Ernotion detection
a) Mochine 8)Mutilingual sentiment analysis
3) Hybid approach
analysis analysis
)Fine qrained sentinent Io) Time-aware
sentiment
Dtopic-based sentiment analysis 12)serni-supervised SA
u)Gragh-based sA
sentiment analysis
Intioduction to advanced
analysis) Lexical analysis
iwod level aralysis (Lexical
(arsing) syntactie anali)
2) syntactic analysis
3) Semantic analysis sernantic analysis)
9) oiscouYSe ntegration (oisclosure integration
s)Progmatbic analysis Pragmatic analysi)s o
Stages of NLr

Tokenizing: Process of beakirg down chunks of bext into snaller picces.o

Tokenization is the tokeizatioo
pTOCessing pipelinc that begins with
Spacycones with a default
a Shap:
making his proces
wovd tokenization : indiuidual wonds
don into
breaks trt individual Senter
)woid tokenization text don into
sentence tokenization breaks
s)
 NOInalizing words :
a little moe complex, than tokenization. zt entails condensino all
'Aomalization is
of thab word.
forms of a word into a single veptesenbabion normalized Irto eath:
Ro instance, watched',watching ' , and watehes" can all ba

Noimalizabion mebhods :
uord
wod is cut of at ibs sten, the srmallest unit of that
Stemming : with stemming, a
the descend ant words.
fron wbich you Can create
a data
Lenmatization seeks to oddvess this issue. his process uses
Lemmatization:
ot a word back to ts simplest form, or lemma.
sbruc tuye that Yelates all foms

Texb classification JMachine Tiansbtion Text Genesation:

Specch to text "Text Surmmarizabion'
-Sentimert onalysis
Text of speech semantic seaych
"Resurne analysis
*Languoge trarslation " Autofilld
"spam analysis
" Fraud debecfion
"chatbots "chatbots
voice bots

ehal lenges in NLP:

)ambiguity s)Multilingual processing
a)context understonding c) grarnmat vaviabiloy
) Blos n data
3)Named entity yecognition (NER) s)Real-time pt0cessing
y) Oata auailability 1o)Trony ond sacasm
)contertual words, phrases d hononyms

English part- of-spee ch (Pos):

- onginal brown corpus USed a large set of 89 os tags
tags.ot
" Nost common in NLP to day js the Penn, tlee bank set of 45
these slides
* Tagset used in
context of a farsed Corpvs
e Rodied from the biown set tor use in the
Cie. tee bank)
used for the brGish national corpus C8Avc)bas4
The cs tagset
 tokenizatien:
Text data as tokens convert text tio
ML model
numbers
"she'is', 'good"
vectorizabion
numbers
Convevting each token into

vectoization :
o0cUment #2
DOcUment #1
Radhika s agood person
He is a gocd boy she s also good
Vocabulag - is,person,she,Radhika
a,also, boy good, He,

ferson she Radhika

also 6oygood He

3 5
Index

CPos):
Frgish ast- of-specch a.pab-of-speech markey.
senterice with
"Annotate eoch word in a
analysis.r
Lowest level of syntoctic and wotd Sgnter serse
disanbiquation.

for subsequert syntactic parsing

"It is useful
Ram eat mangoes
For e xarnple : NNP

TF-IDF Vectoy
Invese ocurneb Frequcncy
Term freguency Hoü cormmon (or uncommon)a
a wod
measures how impostant woyd is acvoss the dataset
docurnent
IS for a

TF - IDF

... inohes two computations

 statistical calculation fov TE-IOF Vectoy:
Frequency of the wojd in a Qoc
nc
Total numbey of uords in the 0oc

imporbart a uord is to the document (without lboking at other

TF Captures how
documents in the dataset)
Num of 00cS
IOF = lo9
word inin Num of 0ocS)
bo distinguish docurnerts.Tfa word
Can be used
zOF tells us if a wod Cfeature) close to o'j-e. qie bw
documents bhen IOF wll be
^PPears in majonty of the
eightage to that feature.

Document #1 ocument #a
is agood boy -she is also good Radhika is a good pe 1aon

Radhika fNUm of 00cs

He
TOF = lo9
word in rum of docs
a

Jood
IOF(He) - log(2/i) =o30/
gcod
Person
TOF (good) =log (/2) =o
boy
she total TF CHe, doc#) =/4 o-I|
also TF Cgood, doc #)a/4 z0-22
total
vocabulary r
a,also, boy igood, He,is,person ,she,rodhika

also boy goed He

erson she Radhika
3 5
IndeX

O.o331]
DocUmert #

bcument #2

TF -IDF ecto
 Seque ntid model:
Intvoductien:
efers to machine leaning modek desaned for data that follow a
Sequential leaning
Seq uence. matters.
clips,vldeo clips, and time sevies data, where oder
- This includes tets,audio ordered, unßke tradibianal nod els
data that are inhevently
Seguence models thive on
that assume dato. points are independent
pispesing and analy zing 6equences ike sentences, time
-These modelsaYe adept at
Sevies, and discvete sequence daba.
sequential model:
CNN matel VS images)
netuotks CCNNs) excel with spatial data (eg,
neural
" twhile convolubional data joffeting a more effective
pptoach
models ave talored fot Scquential
" Sequence
for such data se bs.
identically disbributed Ci-i-a),
and
sequen data is ot independent
between data poinbse.necessibating pecialized
dependencies
cieates
The seguential otder
models.

Agplication :
speech and voice vecognitonis
mode s ale pivotal in
"Sequence
. Time series predicbion
Watural language processingCNLe) is cucial.
*
whereunderstanding bhe seguene
seuence nodelling : output values s g enerated fom
seguence of
"SCqUence modellng the proces ushere a'
values.
a sequence of input
incluing tirne-series data and Gext segvences.
"Tbs utlized acvoss vavious data types,
connected or dependent
datasets where eoch data point is
to
"qUential data vefers datasets
poirte wthin the Same is Cruclal foy
on othe seguence ot otder of the dato points
interConnec tedness means the
Ths
analysis ond interpretation.
 exampks saqence data :
speech vecognition
vdeo activity vecognition
Musicgenerotion Aame entity vecognibion.
senbiment classificabion

ONM Seguence analysis

.onslabion eds avn
Offerent squence modeIs :
- RNN and its vayiants based madels
" Auto-Encoders

seq1seg

Intyoduction to Recumenb naval netuorks (RNN):

"Rw stonds for vecUIYent neural netuork, a specialized type of deep learninq model:
"oesgned for precesirg, sequentil datajmaking it deal for ratual larguage pracesinglwu)
and time seies forecas ting
"nlike convenbional neuval networks ,RNNS poSSeSs internal memory, enabling them k
hondle sequenti al input effectively
.This jnteinal memarg allous RNS to remernbey and vtilize past inputs in cuvrent
procesing,a feature not prescnt in stondard neural nebuorks.
RNN task variations:
one-to-one: troditional feed-forward archecture with a sngle input leadinobo
single output
"one to-mang ;Used for image captioning, here a sirgle fixed -size imoge iput generotes
a sequence of aords o phrases.
Hany-to-one : Aplied in emotion classification, procesing a seier of ubrds or poragrophs
to otput a probability SCo1e indicating specific sentiment

Nony- to-many : for muchine translabion, ike goagle baslate,converting vaviable-length

sentenes fom one languoge to 'another. Also ermployed in video classification, analyzing
seguernces fioone-by- frane to cateqpsize vi deo content.
 LSTN:
memory) cells matiy the traition al RNN archibecture to improve
LSTH( long short-bem
mernoy vebention of inputs over long durations
to be caied
key features of LSTM: allowig irformation
hidden state,
*Mcoforates a cell sbate abng with the
to futre time sbeps
significantly enhancing the abiliby bo
dependencies,
"Capaale of aptuing lang- rarge
ertended periods.
Yememnbey previous inputs for gates) to cffectively marage
and
(input, forget, and output
"Ublizes thiee distinct gates
manipulate the memoy
RNN tuth LSTM:
ho (h

A
A

Applicatiors d impact : leasning for nce modeling

techrique n deep search,L$THS
"LSTS ae aprevalenb
Cke apple s siri and google's voice'
real-uorld opplications these technolog ics,
" foweing tothe sucCess of
cortribute significanty
 Aubo-encoter in mchine transhtion :
acbive ctea of vebtasch
ocus avea in NLP: Machine translation ( ) Temains a highy
ibhin nabural language processing (NL), aimin to develop ahorithms Copable of acurately
and saifbly bansloting text fromone languoge bo onother .
" Core avchibecture :Fncode-Decodet Model. 4
context of bhe soulce langunge sentence, compressing its
Encodermnvizes the
nfomation into on encoded epvesentation. language,
encoded daba bo construct the sentence in tho target
"ecotey :ubilizes the
piece by piece
,Encader pecoder

output :
Irput Reconstyucted
input.
empiessed
data

Seq2seq madel:
seguence of words (such as
CoYe funcbioralby :Seg1se madels are designed to process a
and generate a Coriesponding sequene of words cs
oUtput
gentences) as input
vecuent neuol netwok (RNN)achitectuve.
"undevlying technology :uhle based.on the nemory)
models often employ advonc ed vayiants like LSTM (long short - tem
Seg 2seq
(hated RecurYent units) fov example, ubilizes LSTM.
o GRU
madels opevate by considering bwo inputs ab coch step :
.Mechanism : sgseg
the user.
* the cuYYent input fom
reused as an addtional input.
feedbck frorn its previous otpub, uhich is then
Encodey

Decodey
 Seaa nt atoencodes:

"Seq12 vs- AutocncodeY ; dffeventiate between segseg models

technigues, ibs coucial bo despite their similarities,
uhie exploving this
autoencodeIs,as they sere distinct pUrposes
and
domains (eg-,english to hindi),
" Seq2Seq models : diffevenb inpb and
output
madels work across their sbrength lies ih
Seqise opplications
ideal for machine translabion to anolhey.
making them langwage
one
Convevting sequences from
output
within the Same input l
" Autoencoders:
seq2se9
models,operate
reconstiuct the oviginal input
Autoencoders,ousubset of designed to
lish).they ate
to eng auto-asociation
.
dormain Cegy english functioning through
version,
forn a corupted