LANGUA GE PROCESSLNG

NIP- NATURAL Page No.

Date

Ps a baancl f computea cclence

NanUsal Laguage paocevfng to
A wwelu uakeu a macluine capable of
specmeally of
aumau (ouguoge aud eutncde outpus au pea
wndenteading
NLP & coWsideud a subsef of Mt wile NLp & ML bofh futls
4P.

Need of NP
O Human computee futerccti¡n to e t
futnoo Actaival &exingchen () text Summazhin
Seutim eutal Analytis augoge penenatzm.
language aan slaren.
tet 30 Spee cl
enablisg.
Comput er to uderktuo, interpreat aud gentare
lasguage in uay tat fs meauibgful aud eonteutienal
sevent.

Components oF NLP
NaHUsal Ilanpuage paoceué

TTanguage
Crenentuson (NuG)
Moapholott
Paoguamate Naruha
bscounse. fect
 Page No.
Date

NLU is Pndeels of aLAoli'ng cucf id enpsitng tanguage, pnoduees

-féngaene eutput fnna hunan aguge.
non-lingueio
ugating auct genesnq lanpuage
eutpus pnom non-lnputting ouput,

vtuol assisteucaC rasves).

-Nacline trcnslahe.
Name ewtity ae copulrahon.

queshta answthiy 3ytem.

O phonems. - phonemes ae smallesh wut or lanpueye.

Ex ? vorce, ext, chauats

Moaphemes - Combinatm of phonemes oA chancs S

ex3- oonds (Run, Fungui).

Lexemes lexemes o e foAms of mophemes.

Syntux - combinahon off mophemds aud lexems

Ex sentenee (we'cu

context
 Page No.
Date
Blocks
Blocks 4pplicaen,
SummaÜZOHOn
contet topic mcelellg
(meaning) eentfment Anals.
-Syngax
(sentence. /phAaSe). salaon extnacon
MOrphem 3, (exem s TOkenizut'en
(woads)
shoer embedding
phonemy speeeh’ text
Speaxen deutticatiem

NLP pipeline
>tertpeedu
Data Exnachon.

tex t

feute nng

Modeling

MOel
upd otlen

Dati cpuirton i- It involves collech'em Raguned

fnom aWos somcen.
 s
lemmtiohien Pos
chuntdn 3
emnit
Name at'en ?okeniz
NLp Stof
eps
Aeeipaoeel,
gaunk mewn, xajaival. (nfoauuanen
nemvngifl
eoaupucislon accwaey, clossbteat'oni
eNaluetien (e)
&DL) aPpooel
( bcyed patern approcl based Rule |O
nuode buileing Appacthes
en
on
oectve. wuilize auel modl MLctpnodues wneie
Reecqmihion) Enhtyi NOme, NER( jaghg, poS
hsab nomaliO'n,
tanng, Aumoval, toweniren,
ondsopw -
tosext. oRiented
eohiAssumjtiena
d (genainog
daty) smentunn
Data
ng audpotteus euhty fd uppig fext sca -pdf
toweb
(suvey) "
data time Real-
addta Publde
aaRelaveut R)EDA- doo: collect to
wauys
Date
No. Page
 pos- bae Auns.
tag pos- Cuye aulbe alor tapos
-
Verb NOun
pamnahie seence elps entna
SHauctwe
aud Syuache
ecueleasitrig invovea t togiv Pes
4
3trulan uouse
Mice
Rumnivbeterg
Moepholonièal
Pnpo. Run
temma
auatyris
wsplologicel
ingutiely asyuee
uOds
to devec aeduce to use Lemiizui'
is it cn
Ste. - hoond
common of sh acCtot
tue iuto wondwuile
on end doun
a tUOks algo. Tw's em.
tokeni'zctien, w0nd foKens, sentence
te) coxpus of
(omalet
it tokend uto
Date
No. Page
 Page No.
Date

peep (eami yp pos ttiug

Name Enity Re cog n'tion (NER)

2t nvolvu
fdeuhbyg aunol claamibyig
futo p- debined (n te
cutego Mes.
Tue puay val 's to etnact ouel
tnn 8tNuere eot o ab)etuis name eub't'es
enable weiellue to wclertoo e
coutext nelohenuip betn the eutitie
) Some common
on appnoche of ER I
Rule bese N ER
was Foned in |992 a CoNDON
Deep leang NER by abe.
Shasthccul ER
LoNDON- GPE
1382 DATE abc PERSON,

chunwing (t es gecluique Hat luvetues wOrd s uto

seuseuce iuto clu bue on he pos
ucuuty there
the t.
phaase
uowr pacse, wAd phease ( np, wp).
bok
Loudly
pnoceomg Ashens to wounipuans aud auclye alouhne
The qool 3ext pnoceiyg is to eognat weauingful to aud
astats fom ustaucturs lura (text, en abl
undeeue, fuerpnued ouud Jengce next.

NLP Pnoc eti ng

it susese ony t auayns
dqeuesann to wdeewd umau
 Page No.
Date

8 wey seps foa. pnepnocevin.

sepmentcuhien oreninaon
70kenixahon 9 slopuoand Removal
Noauat'xhon G)E) s+
Stemmfng (6) lemmatization.
Naie
pos eutity Receg/fion
Spelling cOA) echon, G t s an
hical'y peing speiting eror
autoutr er tulu te tert. tt s
CAuel for. Rupnougi t aceua of Np applicahon aud
eukauce ovenal uer expei euce.
leyenshtein
htein Algoaithm /Ed't dist algo
?t is One Df tuOst
One of
cOSlEeh ons.
populcu appinoaclu poa tui gpeiug
The Ceuenghd ein algo. perboselk in iye bas te
S a es -

Step0 : todeuity eucorect

iuco ogols,
?nput
peenagle
App fokan' scuhent 1 have
Pncome ct ord peenapple'
pechapple
: ccud 'olate
feuenation.
Cenera'on Genercut e ser of caudi'cl ate coNECHUn fo
wi'speued woond. tu uts step te possi6le.
CeSAtaion
eonechen by auoug peaahism.
peenpepoeapple
SepApply edet epeaatloyn e. tnsentin, defere ud
subsHf b'on
peenapple
’ 'peenpple--’ deletion
plueopple ’ Subsiturion
pfne apple aser
 Page No.
Date

Stepe cuculCLt e olit

olistCuuee.
tevensutein d'stouee esaatuuareel foA eeel o
Aeueser ed foa Eueorect 8peuPuf. ceudiodote
ED (CP) :3
ED: (C)4

StepS: Find ost Tively cOreChon.

select cc o'ol cte coeien' ih ouoet
evenein disf.
at vely conechbn

cOwuponeu
pnocekiÇ. tt es Stotitical toe/ tat cualyre parien oE
Vauous
NIP appUechen. Buelo au neulina gnanstabon,
speecl aeceg sext eu erath'em
ehanien auc niose..

Stoteroal
RNN
GRV(

Esponeuhuf.
Ture aue two
pe of pnobability. e., Tqué pnopabuity, condhpinag
paotait
He is
eVeut A: He s |eshg S
Veut 8: He f's eafg.
 Page No.
Date

oondtional poobakility &- ft

6 DtDye ue happtdg baed on prwiouy'
mect
caleuwartou, paobabilityE
alauady
X P(B) p(An B)
PA)

) steps /levels of Np

$i. NoQphologfoal Aualyes

3. Semetic

$s. Dis cosse

down wogls iuto tue gmelest pammatcal uwks

known
Nophihen
auc ael cuh'en.
aud deteuuue how ditt uOspes
twoRds. eDubiney to po
Exe

wonphin- happy
 Page No.
Date

Smoaloteal enfo (luappy s adven6 feoam

plafYy swoapheme-la

etattwe of Pue seuten ce to wcetucl He sel aton betn wDgds

auc ow coubine o ecuaiefut snt
Shtehee.
tea foaual

passe taee.
constitunt2 sentencet- The cat Saton heMat
dependeng
Mat
The sett
Set.

- N

Mat
the

COeC then
Semeutc 4halps it cheKS gedence is Syntatiealy
t cheuks t e seueuee s weoul
Senence - The apple eck a ban ana.

faenphets gven tert Eufo. pnom

puvious Due.
Ext- pl eage swttu of pe
Tlis s a taguext seeuee.

Discowse Analas ebobeet of preveous se

t decals ith He ebóect
on te uext geut euce covarlencton,
x RQrn is a Govd studeut. He sualtes wey. le
 Page No.
Date
TOPIC language uelelling

O MoQkOV pnoperty
MOakoy ehaun & H ucthemotical wcdtel ie., ut tse
Simutate aandom pn0ces ocwing ove adurcton
of 4nansih'nd Gern thern. Tu's time.it
ane pnobawiishes onicll tuplies Hu possibiliHes. frcnsihong
pRom one stau es to .cuothes touet.

sunny cl oudy Ralny

Sunny o.8 0.

Qoucly 0.4 O:2

Raing 0.2 o.S.

Rain
o.3
Po
ot

o.t
2 Maukoy chain in NLP
key concepbi
MoaKov pnopety 72cunstten uata'
3 Em'smon pnobabi'l'ty
sanou distaibahon
wodel.

S Hdden MOqkoV Model

HMN q Stat'sttocl uuodlel fe.,
languege aucl
wocline acuslaon auol ext
task iue speeol ~ecogniHos
Maoy
 Page No.
Date

Teain'ng dato ' 7eMtng ata,

Pnanali tekes NLP wjll wil google NLP
Cun

Ram
Jlues

Stepi- Appy Pos aggiv calculaing pacbaaitiby

Srep find cmfston pno balaility ( we
foa eael wuique
M2 qcopu4).
Pranal 2/0 M: woelel vesb
likes 3/.
NLP
n

Crocqle
wl
Ram

Step() Find jaansitien paobcblik ,(it P'ndls l

one stose bsecl on pnevious one)
M
C
3/5
N (0
M: 2

paatabi'tiy oue stofe base ou

phuvious 84aje.
 Page No.
Date

paauat tues NLP Ram

Goepe
Ster
stepS) i- Aprly pos
toqg tet
No.oË coubinai'onsa (No. of pos faas\no. of w09ds toaaty

81
[oil oitl google NLp] eut data.
NNNNE
1

S
2/5

fnd out t hiphest pnobablity fon alove combinahon aud

actoadBug
StegG
vivnbiAlgoaitom
algothm is a dynQai't pro Quuiug uodel sed 30

obsevaton
 Page No.
Date

wil

(N

wi! (pOogle (NP

N-Gecm' Model BiGRam Model)

7iove wacuiue lecuu'ng.

Macline leanulng is jascinctihg.
new dechuolories
tep- Apply Towenizarien.
TLOve'' macline) ^leanui sg'l
EMaclu'ue',learuwmg,is' Fascimcuhing'I
leve' exploiyg'new,'eclnotegres'T

TCrlove), ('mlove'macline) (uaelune'eaurut

CMactue', teauuig'), Cleayuing'1s),.B, fasctiucatihg )
[Clowe), (tove'exploig), erploaiy 'neo).

cospus
 Page No.

love 2
(love' macline):
('al'ie, 'leaiy):
MoLhine', leanu)
sfinatin) fascinatigl
(ove' 'exploing)e explondn)
(explcarng ne:
('new', seclnolefies) t.
jelmoloies s1.
Step (4) problsiLity caleulatem.

paobaloLiy of od given u
olivisled by
P(wnlw-): Couut Cwoi, wn)
couut(wn-)
wuere

eount(r)
PCJeanidyg,ove): 2
o5.

PTeain,Macliues)= 4:
 datci
teaddacss meoole!
to
edalcnuage, uachtw'sm Smooting
couudhy) oue winise
of psine, the P
ell
odPae coaty either choose
0sdove'
yat
'exploaig)
:o5 efveqsauu bi the bcH stuce,
(love?
o.s auliue (ovef pnobae'lity
(ove'meLelu)
f
tve! 'th' ts that
gaam bir gdeutitiy
theo
rove atter yoncd 'next Preciet
te
2
matl pejstwosd
agte pcehusecUdg
phe as tocte
, conpus,
e ocewauee
He de eahuy) one macaieene del
qdeutt'sieSnce,y
a6temocliue, ueCtwar te
Prediet
Stefspnadchig
word. next
No. Page Date
 Date

Mee pnoblem encounderiug MSeen 04

to
Dato Gparuity aebers
Rebens to he
fato wiel cela teads
paobalaility esteenm,
fsgueut i y ald
uateeble
paokatiuiey
Sngothig
. taplace smocthig
koeses Ny
(D Additve smocthisg katz Smoo
Backof tsepolaion
oue/ Aco smootug
Laplace Smoo'sg se siaply add f to all Hu
Fn
4everge q x§o value

tue cat uSat

tue ca Saw e wt

Sa fue cat

cowt the'ct)
ceuutet') 3
couu (sa', u)
comt ( on', 'the') 2 cowut'on'):2
3
eouuohe''mar) Couut( a) 3
COut souo) : 2.
'Sous', the') : 2.
|cOwmt'Hhe'doa') = 2 couat^oleg):2
 Page No.
Date

Cowt ( wn-)+y
payatin coapuk
apiy aplace Smeothig.
P(atthe): cowit (he, cat)t
couut( he) t
R cat sat) out( Sct,c)+!
CowtCsat)+7 t12
2+7

Plston) cowutfon,sot) +1 2f/3

ouut(on)+7 2+7 3

P(on|he) couut(the, on)+ 2t)

couut(the)+7 S

p(the weu) couut(uct, He).a 3+1

couwtMct)t 3+7

cowut(sauo) + 2+7

PSau tu) co ut( tue, Sauo)+) 2+| 3

8t7

2+|
ccut(dcy)+ 2t7 3

Fdogoa)? cout (eu, do)+) f+!

cout (oat)7 3+7
P(dog]saw) couut (auw, dog)+).
Rule based StochasA
,144Afoau dhiou based tagig
Page No.
Date
find tlue bi:gtcuu FOR teshing pwpose,

testingcoapus the bine sat on 4aee.

BÙ:au& couut (Hhe blal) =0
cout'bind!eu)
couut ('sat', on') 2
eount on'; '49ee): o,

Ahd paobalait/ty ustug laplace smoofi

P(bind JHh): couut(He, bhd) +| Vrs
cot (tue)t7
P(&t| bhd) cout bincl, at)+l ½
cout (Ot) +7
ouut (Sat, on)+
cout (gct)t.

cout (on) +7
Benebits

Unseen

Reduei ovenit
imhlys peesltehen
biy uakes alicl wcdel leaaikg
te SupMsed class6 CUe DM samtnc'ye
Mocleli
io. ofnt woolelln precltehbu
by tuues
wawes
lel
MLA clause Qauge wide cne S- Gennatve
eM Moder
(roint)
)) MOde
h've GenNtive
pnobablist'e
TaansOnaon
base
Tger. Based
Stoeaste 7481n. Based Rule
TORHn Types
of
Mbdef. fiye Disafui (2 MOdel cuhve Genen U
sORina pos os, Usecl cue ouel Modlels two ae There
bots chat
evies
oewndestaueliyg nnslahbn, Seutenie
lyos eic live dota seueuce wth wonk
oloility
tO
quece.
Tle activity.
video fruslahen
cud classibie
cubie Geutiluelt
rahby sieJecoqu'
"Mu tion,Macl'ne preetieens
tines ficunelo fhatueltn chen appe vvety gddns
o
toeseol becau wuodel csequen
ed Jeonuing, oupev'sed
Teggn cl4betmf
Date Mode
le
fosequeutiel UNII
O2 gulyu 31
No. Page