0% found this document useful (0 votes)
7 views62 pages

NLP Module-1 Notes

The document discusses various approaches to natural language processing (NLP), focusing on rationalist and empirical methods. It highlights the complexities of language comprehension, including syntactic and semantic analysis, and the challenges faced in capturing the nuances of human language. Additionally, it addresses the importance of knowledge representation and the limitations of current NLP systems in understanding context and ambiguity in language.

Uploaded by

Rithu D
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
0% found this document useful (0 votes)
7 views62 pages

NLP Module-1 Notes

The document discusses various approaches to natural language processing (NLP), focusing on rationalist and empirical methods. It highlights the complexities of language comprehension, including syntactic and semantic analysis, and the challenges faced in capturing the nuances of human language. Additionally, it addresses the importance of knowledge representation and the limitations of current NLP systems in understanding context and ambiguity in language.

Uploaded by

Rithu D
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
MopulE—1_ Chopfit—1 —Tekohcton et ee Shee pores mam er Communication uated humans. Hy the tool woe wre to eps te gtealis park oh oud voter anal ewchons. Hy phages Thou, hos a thuctuee “are i scape ated ong CHU) ip Contermud vot tee clereloprerl ial emodals oy aspects tf, human gee my | Tr -hoo main Yannis 48 clawlop rant i= A Develop arstowatidl tool ak lenguoge proto 2. ain betin UES a hueman, Communitation. o dompelonal modal willy human language Penis aleltins weg wiser oO Mroislitas howd. hu SR. ome) prow lenguage t ieee te: Two major approautus ~b NLP. 1+ Rosionalit oppoads - 2- Empinvaxt approach. CHETAN. R Asst. Professor Rotionglnt appradls Whidk auumes The epictonte o pome language fovally in) hurwary brain. Supporters of tes aapproatta ongur hod jx cnot posible fot childern to loan tompler ‘thing Uke vredural language -feom lachal 4g punsoug aay cia) ie Thy clos mot believe in tre eievetente. df Rome. lompaage fonslty “Thuy believe ty He exirtente % Avvia. gprusol Orgamiahio principles fuuh od patten “retap nition girusazalin ama ampuation . Leas of. cletarled shuctutes com , “tnaeafees., “take plow. Tutough “Hea applitadion of toe. — on Sensoey ipa availa +o the child. Onjim of NUP Noha! language pouring orator, eemad, neal m teamalation npg wang pl fom rin “xkconth - Nodueal laug wage ratio aban only ihc 5 ott Notusal ae preva indoles both 7 © Compicstional lingua i Simulag do thecretitel and Pes legions Theosetiral Ling uishig : “Tay mast povide chuchital destaiptron % nedusa! language and th gewartite- i oe mot Coneaedl with toe avtuall pocwing of Sonfentes ov generation of Sertentes tom churtucas alessphin. ate in @ qbeat BK prnuples Hot xemain dommon autess laugucger ard tent vals that coptute inguistic necalzahin. Ecaruple most Languages hove concuts like noun anal Vewb phrases. Theosehtas a idanhify wile, Heat dersvidbe anol Nett the ee % rg uapes 5 ee Asst. Professor Pegehobinguisticn t They &yplaxin how humans prolate anal Conpreband natural language « Tay aga. intersted in “he veprenenttation of Ligue Gruchutes a6 well as in The pron by which Ter chuchites aga produtoal « Thoy ey priverily On eupiviial investigating sto hak up “bees Heons. - , Conpidoshonall_kinguistey Hj} Conurned with the lan gu uaa tonpidehonal model of ie i. a Shak wih the applitecion of Ligguishe esovies amol toapedalad Fechwrgpies + NLP. In Comprelachonal Urguishin , reprenrting a language hea nwojor problem ; most Knowledge. epson darkh. oa a Small pact of Hmowledgen. Compudehonal rwodls mou be broadly clanified ais Know! deiven and data-driven _eachgonts. ie ledge diiven Suytims vy on ex ply edie Lin guiatte Knowledge , often expend aa wt of handuxafed gtammak Kiales , Acquiring and tntodivg futh Krow ledge i diygiudts Dota driven approaches prove. loupe amoud rf ala and toually employ tome rmathive laontyg febrigus to ban Sy route patfeeny. The amend of human oper tA ley ama pesfeemare | ap thee tystions aliperclent on The gpertitg of teow data » Mou Sip lnns ate wrwiably adaptive. do nohy data. ® Langtage amd Knowledge > Lewguage pte meokiom 4 expreaion ww whith fools h olesplaced. Hawa uage: 1 Pes a wuedium % ex previ, jp he cues feton of, “te Content it Sxpreses. “The tome Content “can be enpresed in differ laupua ges Te apeesia of one language A waritten in the Aame ee Faced 1D Leweal trolyus a H pe Smplst level, lat involves in rays a wordy. Word Me the mrt sfordamerted unit of — modural degt. Word lave? pomieey voyprres morpholog Knowledge. he. Knowledge about “the chuchse anal feemetion % words from basit. units. Ta Rules +" fog work ftom moyplames ake l Spouger- alla a He Souste Code % @ Sheeam id Chasarters anal , comeals into macanivg fl lexemes. Bt divicles “the colle ext Into poragsaphs , ferences aud wwida. bb Syntactic Frolysi.. B comic of Sequence of worn aa tenit, usually a Fertente and fing ity shucute. Be decompaer a Sentence into de Constrtuerty aural identifies hovo “hey vila. to cath ofan. B Captited danenalenly or rnon-grommaticaldy Cr senten tes tookivvg Sg tonstrounts like word odes , numb and tre agiseemuct. Thy lvl of pees neypiews gyrtauhie Fnowkdec, te. Knowledge about how wor axe combined ay a lager uni purr as prrasee amd Sentences, For Sxomuple, TL wert do “te market! pa valid fertente lures wert the T ynosket to) jr note Semantic Probie) le * Cumantin amoutted vaflly Hear meanangy of the language, He jp Concerned with creastg esamivg fel “eprrenttadon % Unquiohe impus The oprueal tolea, ci Semonht. nteepeckodion in to take madutal longurage andentea ond! map “Ham onto fome. ea gel afl D meat Covpone jn digger an gon Ns orion Os be mosaning lou. Ne We rena pov af Semantte auraly sis ie here eeen Pervardig. A word cam have a rnumbes of prssible. mmeaming amotatd with it Buk ina giwn Cortert only one of tae rags padiuipatia . Firdivg Or he cowedn vita oF 4 Postuslag wat word fe ne taey to find toning nibs CHETAN.R Comsioa The folauing Sevtemees Asst. Professor Kabix mel Ayan ate momied, —— (a) as Kaby and tuba Au omamied. Both Sentenus © hove teadital Chutes aol te rau whags 4, indiviolual word ate Chas. But must of us end up voith soo effecond y eapectoshine - We oy inheprat “the petond fertone. “to mean Tat Kabir Sed taba ate raattced “to each otten, but the inteapwetoson door not ecu foe tee $ont Senlente- Symbaht chute aud Compo sshional Sewamhys foil to explan tho. intteepeetatos - We make use Of progmahe vali n. (2) Dicovype- telaea & : Dycourpe— Level poumig abe to inte prot Hee chuduee aud mio ever unik, eg: at he pesogiaph amd downed 1 in) term ca words, phrraney , chuvshees eel and Lendentes. Be vegies the sesoluhion sl amaphyn'e. vefdtonces ond iderdr{stottion of Facoute chuuge, Be alr ~vequives Antcous. Krow idge, thot th Knowkdge % how Te | oy OX pewdente. cleteeniviadh bY prewali evident { Th fet, pragmatic Kroubde roy be msoded -foe eee [Link] wepeenten « for example, in toe tales fortkentn rmescluing the amaphore: "vefetenue Ha veqpiven pragmehe Know ldge’ "The dott adminishahon wefan cto give. he rade. union permatesion ft be wactng betouns. ‘hey feted violinte.” 8 =the dbistwd oudministrahsn, “efured to ge te tade vndn peematssion + tun mading becawne Any oppyse. gree 1© Peogrethe, Probe: Thin tre baghot lena of premeig vouch oltabr wotly the papenepel uae of Gardens, in SAhuachons « 6 regree Knowledge. sem wodd, he Krouledge tot extends Laid Contents of The text © The Chollngey of NLP coerce There ate rumba of; four tak make. NLP youl, Ot dak to the problem of, ~epeenerdecion and inte pretahsn - Cor Treg ire, preLare “reprountahen oot a \ r a are Wally aanbiguous and vogue, athiving futh Yepruuntoton Com be alejgeull ’ © The inabildy to caphute all he sayuteed Knovlsdge. ix andl four of, digrually Ft fe abmogt repos to embody all Sourtes of Knowledge thot luomary we to pret begeeg @ “The qetert fou of epg in nohual larguoge ip iclonbepying te Semartign. words aloe clo mot mahoa furknte Tread, it jy ta words os wll as “hai Syntactic aud Semonhe vuohon thet gt mraohing +o a femtonte.. 4 longue. Koups On Sc Neo words av adolsol Continually aud enphng words ate inbrduud in ned Cones > For Ceomigle , mech rawppapern aud TW chonnols Ue fu to ops to the Jervowiph att en Jus. World Trade Gute in She USA tn Doo ® Bioms, rmebaphor awd aLipses add me Conaplens do idatihy us. “snips of, he covition text. AS an “ereanle Conpider tae Aetente., “The Od man rally Kroked the bucket" The. moons of “tin pertnte har mottivg te co with Aa words "Kick! ond “burket? appasing ue © Suara ape je omotteh problem. “The sope of qpantigiers (Are, eouh, etc), i often ot Cheah ana poss problem in axdomahe pong: @ We re of natural lamguoges h anothe§ itty + Lepital Ambguty: - The fist bel of orbs shy sya ok “the word level. Wrthout muh +ggeets we Can idech ft words that hove. roubtiple. rraamings amounted dt them. Ma x books a math. Sn ‘ above ae word madd Yefets do Avt eth, Howmga’ yr t00 oe pathes oF Manga ip Lonlaes +e 4 mata, (¢ ov cles metclr) NP + VP S> Seocenta ve — V+NP AP Nhu Phrape NP —> Det +Noun . VP Verb Prope Deb > Detecnsius V —> Aux + Verb De —> “the, aan, -- Verb —> Code , will, eat... Noun — pole, “Snatclrer) >> dux > will) fy Gan... Vaauapsevabiondl pene ds a pk a “Pemsfetmation vl, whith hawafeem Ons pheae-makes into anctieg phrope-makis. Thome fihs are applied om “he -heminal shing ly prone, Shurluse “tales. Unlike phanne. chute. wiles, emotional Yules ate hderogneous ona. have ag move hom one bol on thaik et hand cides Tore yu ax une to one Svogale ~epreardation jlo omethse. eg attive Sedente into pamie Ons: The Sule wen ative amd paie puntontes pe NA fuk -V—- NBL —> NB Aux the ten “V-by Nf “Tomah vuls tan be seletors os optional. An ObY reve nasidn ix one that ovputes agteement in numlbe® Of Subjede amd Veo, Dn tonal) remasporahin je One “Wot rnodifies the Srudue % a pentente, hile prsaving ty dong: Mowphophomena'e nul match eat fentenie, ~eprourttaion 40 a wn * phonemes. : cnet sot Consids the attive gerdonw: ps The pole will Codd, the hatches. The aappliadion 4, pleon ute wule will anstgn “the prucite Shown below: S NP aie oe ~~ Pow ae aS oS Bur | y bet Noun The into Wa + Calpat: Be wilcebe een oe Cah by S ee VP as gos powive. roms feemathin rls will converk Le perder polite. eat ie s wil be on Prowning nd ee Meee) QOL SIE Thue of rumbX of clifgptentes behweun Tian bengtuages amd English. Soma. of; thew differents aks Gated byte? — Unbike inh Tndx suuple hae a non—Lingos shuctute. > Unlike Empl, Trdan Vwb) os d t Coddente. Chute. hove. a Free wird odes, fe, words Con — Trdon be mmoyed ae Le ae without Changong ages have. SOV Ciubjed Objet @ > Spilling stardatczechon jp move Subse. in Hinde than in Englah- —> Tran lagu have a ‘velati vih_ pet mrp Vararils . “4 i —> Thdian languages amok exctenaive. anal productive ure of Complex predicates (Ce). —> Trdan age Ue post poirhin (Korrakos) cure workers inglead of Peposihons: —> Trdan ue vb Complexes consi %, Eequnas of Voth. eg Xt eo Sela eee The aurbiawy voulr in tus pequente ide. i ron a save gt ol a ies is NLP _Applaationg —_ The appltachons utilizing tthe NUP intlude “the folowing’ A. Mathine “TFrmslaton: “This fen to avdomatic -romslation of “ext From ona human language +o anchhev. “hy order -to con out There reomelahon » tt jx Necessary 40 have an understand % wor and phrases, gyemnana of “the Awo' languages wwolved , cemartha of tee ju owd oe roledge. % mG t pnt. & ect 2. Speech Rewanitor The ix the proces af ppg atoushe Speech, signal to a set of worn “lhe liggiaultin one due to wile vorwehiorp in the Pronuncaahon of uae , homonyn and acoushe ombig icin, 3. Speed Systhesis :- Hy fen he Caiter problem: when a Gousmunt A indeed with ong teem and hae qputy contains, a daar term, ond Aha Ao tem Chaka A Common mdanng. Prise problem ust, i word —baaed vvebrevo ip tat it por Femantic and information in ‘tue ~rebiwval prow. “Thin fr4stmahor fh lart in ua exbrattan of: woe ds fom tha tent and cannot be vecoveted by ha wehieval algorttans. cicaaie Asst. Professor @® Trappropmate chavacteryahin wi by ee der. The Une ray fail -fo intlude wel term in Tha qpery ow include tovelevart tevmg. Tappropwahe ox tracusak, ques rad ‘bo poor vebaeval petipemance.- @ Retevy yy wreprarttahon vith that A she dowment w another eu. @® Seachon of the appropriate Ainalagtty mogguse. jp a Cranial) put tn thee clown Of, TR” cyrtom. (5) Evobuating ter peysemance of TR syste jr abso q mma} — Recall ane Ho mt mat © The mrajon a oa of TR weto Search a doumunt Tn a vrarmed ae to “the ara 1 Understanding hak constitutes “relevance. jp also an inpoctomt japue @ Wa & dourmant collchoins and he Vi | nude abe alo compliate. ext “rebreval “oa | MopuLe —1. ©. Chores 2. Language MODELLING A model Pa cluceiption ok ome comple entity or poun. A Langage model jp nus & duixphon of ieee Trdeed , nocaal Largaiage ha Compl» ontity omd in order. to poaw it “Tesoughn a computer -boared pogo, we mcd to build a ‘eperertatin oy it. hh jp Known 04 Loma ucige Wadleling | Lomauage modeling com be Viewed ert? a4 0 prob of Faremoas weyers or a problm o probabilly estimator. CHETAN. R Professor Tee one 2 approaches +f Lenguage Medelivg: salsa e ipa eS De oum ctu qremmay of a language -bo crate ihe ened. Ee cette fa to vepraernt the fetal Shutuge Bf, & lomguage. Grommas Consitts of hard -Coded rules Aefosng, dng Gruchte and — VaxtbUy Conakituantts appearing mo Unga vO uit. fos cromple; OA Senlente Usually Conmals noun phrase om a Neu phiove- The gferrmncd bared approcly odknph eo ublge, Tah Shuuke. anual alro the velachons bos © Slo Lagge Mol TH ited a language model by Teateg it fom a Cexpur- Th oxder 40 captuse Geeta er a language, aha! ‘hoining Cotpus mad -to be sufpjtierthy lasge. Stoksteat language wodelling 14 The attemel 40 Cah gee a of, Inprwing tna perfetmanta. of Nasu noduead langues cnppltedtons-" Stottistivo) louguage omedelling W one > ha fordenenta! oaks many NLP aapplitachons , ad Spach ‘veangition, \ Spelbing Coovechon , harduiting “Wetagrition anol Marling hromaladion: Noxioys Gronmma — Coped Mod RES Pet aa (2) Genutodtve Gromrrprxp noe Noom Chroms! in [407 in he book Syetackic Shructutes woot Hed ue fon oprusati. pentontes iy a language we Know 0 Gllshon of worda. aud why in thot we 4 eo Fertentos that can be generated the wus ave grammatical. “Thuis point of view comineded a al Unguish and p Called hos pociracipabole. o yr @) Language ik laa betweery the sound and ip "mowing Au omy model oe longuage Bhould ono deal with He aan e “ta Sentence. ai Coe est ® Hiewrebral Grammad we Choms descerbed Classes oi geornmay in a hierarebeeal manna” where “he stop ayes Corttained the grammar, eran 4 Mh dbs Ohisins Hence Type © Cor umashided) geommes Contains Type 4 lox Cortext Sonitve permet) wll in tuen Combains ler (ox Contest kee - geomman’) Quel thet eno Covtas Type 3 gfammoas Cregudas grommet): Coe ee As. China Viewpoint. -taken by Uonguish jn hat the Shurhute of & Lorguage cam be lundeestood at the level 4; itp reaming porhilanhy whale vesolvig Grcduraf ambigul However, the Smbenter ake gion at the Sirdashial ak ama, “tre From seemacion \sftom muamiyg to fyrten or vite veesqa tw not well tn deastood. a | “Transfoverashonal grammar | assume hoo lovely exten te Gontences one atthe Guctace level and the othes ot “the deep ~voot level. ' Government and Bindmg Fheonya have svenamed them as Sled and d-bvel and iderchyied two move bevel wf “eprventobion called phonetic “form and Lagitell form. d-chuduve. | So chuhere Phonehe “form Leg toll form Tras formethonal genes hove hundvedy of svewrih wcules , whith ane greets wage Speeegee and aho consul Speoge- Genexcchon of 2 Complti St ef coleverst vuln may not be possible. Th 6B Yue define Yulee fet trite unis af the deep lal, iF will be patsrble. to qt np long uage vill fs veule.. There deep —bvel ruchikes ate 4, noun- phaare, Veob- pheaxe , ee . Common to all ae “ Let us take an example do explain de and s-shuchutes " Mukeih was Killed ” (In raneformathional gfammat dhs tan be vepuperted a GNP ux VP 0% ee below: Ag po oa oe NPs Aux NP \ Ll Mukesh wa, Killed IW BH 6b, te tachuchte aud dchubhse ate as follow: Ss i & Ae Molesh was Klkd > ie Reeve! RSM: NP INFL. VP | | ne ©) past kill Mutoh ° | Mux pat vy vp | : Pat “VP | INFL: Tiefleshion | ey Ne — : a et er auth kitted © mtQes era > j (2) Surfou hehe Dap Shui. Comps 96 Govewnrverit and bind (66) Compeien oO at of “theories thet “the Shuchyes from cl-shucluce to cehurture avd +o logitall em (LF)- 4 gprecel Brom 4emational meu Called "Move ob” jp applied ot deshuree level os well as at s-chuctuee lvel. The Sivupleot en + 6B Gn be vepryented by below te sides + fora] Move o& Move A : logreal fom 4--~-] Taoria and Principles Hene 9B conink "a seers of modules “that Contain Conshrairth —amal prnupls applied at Vooious levels of dr ~epresentishors ama honmfermational vule , Move ol. The §B Conideyr all ucce levels Yepruertectone (de, or, amd LF) a fyntauhe amd LE tn ato Yolo to ning oy Semanthic- iinkea pretive mmaerronsarnd- The %B applies the Aome Move of tromfeemation +b swap d-lvel to slvels ov s-lvel to LF Level. LF lovel halon » gpomthfien nee awol alo in handling Vosious fentente Conshurtionn Suda as panive. Ov inorragadive Conshurhons. an cnet Example: po "Two Courtrig, ote vipited ty mort tavelleyr”. Th too posible logiad ferme ome: Le: [, Tow couriin om winked by [met troller |] Lo: Panna Move & Lge ra a> ane vpited ty € | TFL, dhe inhepectation jr thet most tvovelleyx virit the ame wo Countrin . Th LF2, when we move [rest cravelley) outside the ope of the Sentence, he nt ecpretadion can be thet moat trowel vinit choo Counter, valet may be dig gecent 4p dierent hovellus. One of the *mportoatt Contepta in GB iy that oy Consheoints . Hs the post 61 -the gone wlathh pobibils Goxtain Combinations aud movenerthy Cllgewine Move A Com move angina ‘to ani posible position. Ws GB. basically dhe fermulacton * Hhaprien ov prinuples veluth boat, Cnshairh to diralleo the corshuchon of ill-formed — Rerthentes. tha Organiza % 4B h an below : 4 Figure 26 Organization of GB (adapted rm Peter Sols 1985) x iz te X K one of te arta) Contepta In QB. Tratead of, cefinng pevcal pheane chutes amd toe Peotone phurhate with Separate Ath of rally. Ye theory defines “tram beth as mow'nal projestions of Aone head. Noun Phagre (NP) , Veh Phsope (Ve), adjertive pawl oral prepoprtional phaare (PP) ate rmawimal projevtions = 17) wf foun (N), verb (VD, adputive (A) amd. prepopstion CP) ~upedively ond Con be ted os head X of thas, ering phiases (whuse X= NV, A, P34). Bren the gentene Shucure ton be “egaeded os the woximal projechon of, frflckon (inet). “The 6B envinages prrjectons at two bevels —> projertion x4, huad ot Semi-phapal level demoted by Ss —> marina) paection at We plropal level denoted ky x The fot level prjeton je denokd as S and Setond lvel popdin ix denoted by The below fi deprti te general aro partwlas ref ¢ i Pructutes with oramples. x €—— Maximal projertton- G) Nyt, We Comides the ~epeperttahin of Tae NP, te stood in a Ahaloa. Tur rs followed by Sue Raprorrdadion o, VP, AP amd PP Ghar. 2 NP: dhe food in a dhaba Cue the a oe [in a dhatha ] the ON PP food Ma dheba Nep_-chuture Shruchice 9p. VPs ate the food in a dhaba. Ly Ta Ue Ie il] fy “™]] VP ate tae food Th a chaba VP Shuctuee ; 3. GBP Vey proud ef = Counhy a ; oy MIL if Mies al oS Das * \ a) Mey Lo oso hts Sia 2 ae an rade! fe bug all [, (3 P oe [ AS in Dek N hood dheba» “Wp. S:ttat che oft tar food in a daa. Ds oe As Lv Ta igs tet» oto] = ow Corp S thot NP INFL VP 1? Seb Sho fat ot hoitend in a dhaka.

You might also like