Luận án Dịch tự động Anh - Việt dựa trên việc học luật chuyển đổi từ ngữ ngữ liệu song ngữ

DỊCH TỰ ĐỘNG ANH - VIỆT DỰA TRÊN VIỆC HỌC LUẬT CHUYỂN ĐỔI TỪ NGỮ NGỮ LIỆU SONG NGỮ ĐINH ĐIỀN Trang nhan đề Mục lục Chương_1: Giới thiệu. Chương_2: Tổng quan. Chương_3: Mô hình dịch BTL. Chương_4: Các bài toán cần giải quyết. Chương_5: Cài đặc thực nghiệm - Kết quả. Chương_6: Đánh giá kết quả - Bàn luận. Chương_7: Kết luận. Các công trình đã công bố Tài liệu tham khảo Phụ lục

pdf74 trang | Chia sẻ: maiphuongtl | Lượt xem: 1724 | Lượt tải: 0download
Bạn đang xem trước 20 trang tài liệu Luận án Dịch tự động Anh - Việt dựa trên việc học luật chuyển đổi từ ngữ ngữ liệu song ngữ, để xem tài liệu hoàn chỉnh bạn click vào nút DOWNLOAD ở trên
nphanquatho, thisekh6ngkhll'nh?pnhangdU'9CmOts6 trU'onghqpmaclingnhanngunghia nhu'ngconghiakhacnhau. 4.3.1.1 Cd Sd CUAVI~CPHANLdp (NHA.N)NoD NOHIA Lau nay,chungtaqua"quellvoi cactu di€n thongthu'ong(donnguhay songngu)du'<;1cs~pxe'prheathutu'abccilam1,lctu,chinhVIV?ymahaimvctu "animals"(dQngV?t)va "zoo"(sathu),ho~c"aunt"(co/dl)va "uncle"(chuI baG)du'<;1cd~tavi tri ra'txanhau,ch~ngco lien quailgi voi nhauv~m~tngu nghla.Tu di€n rheatr?tt1fabcthlh<;1pIy vach~tchev~m~thlnhthuG(hlnhthai) nhU'ngl<;tikhongh<;1ply v~m~tnQidung(ngunghla)vaclIngkhongphilh<;1pvoi tU'duyngonngucilaconngU'oi. Cacnhang6nnguhQc- tamly dachungminhb~ngth1jcnghi~mIa: voi IDQttukichthiGh"aunt"chonhi~ungu'oikhacnhau,thldasO'd~uchobie'ttrong dftuhQnghIde'ntu"uncle"tru'ocnha't,di~unaychungtor~ng:ngay"Winoibell trong"cilaconngu'oichungfa,thl tu"uncle"va "aunt"dacoquailh~voinhau. Dayclingchinh13.n~ntangly thuye'tv~ngunghlatuv1jngmade nhalamtu di~nphanlOpy ni~mdad1jaVaGkhixaydvngcach~th6ngphanlOpngunghla vagallnhanngunghlachom6ilOpdo.Be'nnay,dacomQtsoh~thongphanlOp 105 nhutIeD,nhu:tu di€n thesaurus,LLOCE/LDOCE, mq.ngWeIdNer,h~th6ng nhiinCoreLex,.. Ke"tquanghienCUDv~phdquatlIgonngilchotha'y:MQtsO'phdquatlIgon lIgula tu cac hi~ntuQngtamly - ngon ngu hQc, VI the",mQtcach khai quat, no ph\,!thuQcvaom6iquailh~giualIgonngilva titduycuaconnguoi.MQtsO'pho? quatlIgonngilkhaclq.ila nhunghi~ntuQngv~daDtQc-ngonnguhQc,VI the"no ph\,!thuQcvaom6iquailh~giuangonnguvavanboa.CacnbanghienCUDchia pIuSquatlIgonngilthanh2d<;mg: .' Cacph6'quatv~th1,1'cthE(substance):la nhungnetchungv~s1,1't6chuccac th1,1'cth€ ngonngu.Ch~nghq.n,mQingonngud~ut6ntq.icacphq.mtrlldaub ~ tuvadQngtu,nola cosad€ bi€u hi~nca'utrucchImcuacalltrongmQingon ngu. - . Cacph6quatv~dq.ngthuc:ch~nghq.n,nguphapt?Osinhcoir~ngbQph~nco sacuacuphaptrongmQingonnguthlgi6ngnhau. NgoaicacphdquatlIgonngilv~nguam,nguphap,ngunghlala nhungph6 quatchid~c~ptdimQtphuongdi~nkyhi~uho~ctdicaibi€u dq.tho?ctdicac dU'c;5ebiEndq.t,nguoitaconehu)' tdicaephdquatngon ngilv~kyhi~u,chung d~e~pWi cai quailh~giuacai bi€u d?tva cai duQcbi€u d?t.([29]tr.273-275). Dii tu lan,trong"GiaotrlnhngonnguhQcdq.icu'ong"cuaFerdinandde Saussurediichirahaidq.ngquailh~:ngang(tuye"ntinh,hInhtuyen,ngudo<,m)va dQe(h~hInh, tr1,1'ctuye"n).Tuongung vdi quailh~ngangc6 tru'ongnghIa ruye"n tinhva truongnghla lien wang, con ung vdi quailh~dQcco truongnghIabi€u v~tvatru'ongnghIabi€u ni~m.TruongnghIabi€u v~tla t~phQpnhungtud6ng nghlav€ y nghIabi€u v~tva truongbi€u ni~mla mQt~phQpcaetu co chung Calltrucbi€u ni~m([2]tr.172- 191D. 106 4.3.1.2NHA.NXET cAc H~THONGNHANNGD'NGHlACOLIEN QUAN QuakhEWsat cach~thongTItanngunghiacuaLLOCE (xinxemphlfIlfC 8.1),LDOCE (xin xemphlf ll.lc8.2),WordNet (xin xemphlf IlfC8.3.4),CoreLex (xinxemphlf1l.lc8.4),chungtoi nh?ntha'y: . Cach phan chia cac lOp cua LLOCE tht!ccha't13.dt!atIeD co so 19thuyet phanchia tru'ongngunghiarheatrlfcdQc(tru'ongnghiabi6u V?t va bi~u ni~m).Doi vai Wordnet,ngo3.ivi~cd1,1'atIeDcoso19thuyetphanchiarhea tru'ongbi6uV?tvabi~uni~m,nocondt!arheacosophanchiarheatru'ong nghiatuyentintvatru'ongnghialientu'ong(quacacquailh~chucDang,bQ pMn,tint cha't..). I, . Voi mlfctieubandgu la h~thongcac9ni~mchungnha'tchomQingonngu cliaTItan1o~i,Denvi~cbi~udi~nh~thongcac9ni~mtrongWordNetdu'Qc d11atIeDco so 1)'thuyelv~ngonnguhQc- tri nMn (cognitivelinguistics), ngonnguhQc- tam1)'(psycho-linguistics),..nhu'ngta'tcacac1)'thuyetnay d~uhu'angtaimQtmuclieu chungla nghiencuuv~s11chungnha'tclia mQi ngonngutrenthegioihaycongQila ph6quat (universal)cuangonngu. . H~thongTItanLDOCEchichutrQngdendanhtu,cosolu'QnglUkhalOn (45.000)nhu'ngsu'phanchialOpngunghiaquatho(chico32lOp). . H~thongTItanLLOCE cou'Udi6mla dongian,h~thongphanca'pchigBm3 cfip(chud~- nhom- lOp),soTItankhongqualOn(chig6m2441TItan). . H~thongTItancuaWordNetra'tchitier,d§ydu(chocactulo~ichinh)VIV?y s61u'QngTItanrfit lOn (han 100.000TItan).WordNetco u'Udi6m 13.phanca'p chi tier (hangchlfCcap) va giuacac lOp (synset)con co nhi~uki6u quail h~ khacnhau. . H~thongTItanCoreLex(danhchodanhtu)phanbi~tdu'Qctu dBngnghia (homonym)vatu'dBngt1;l'(homograph)trongkhidoWordNethlkhong. 107 >-TrQngtameilah~th6ngTItantrongBTL la d€ khunMpnh~ngngunghIaeila tuehoml,lcdiet diet,ehukhongph,Hehoml,lediet hi€u (dn e6tri thueehi tierv~thegiai th1,1'e)Denkhongdn phaiphangiai ngil'nghlaehi tiernhu trongWordNet.Vai s6luQngTItanquaIOnnhutrongWordNet,thl ehungta khongth~xayd1,1'ngdilduQenglili~ll;huin luy~nt6ngquatehotit d caetu (dn ngli li~uhangt)1tu). >-H~th6ngTItanLDOCE thl qua tho,khongdil suekhu nh~pnh~ngehocaetu eunglOpnhungkhaenghla. >-H~th6ngTItanCoreLexduQexayd1,1'ngtueaelOpeCibancilaWordNetvae6 caemaso'TItanla caetUviet tatehlid~u,d~nhahClneacTItaneilacaeh~, th6ngkhac(ehidungeonso',hayehli-s6).Tuynhien,chungelingchig6m39 TItan eCibii.nva 126lOpd~nxuit chidanhehokhoang40.000danhtu Den kh6apdl,lngvaomQtM th6ngmaivaicaedanhtukh6nge6trongdanhsach do. >'H~th6ngTItanLLOCE eokhoang2441TItanehoh~llbetcaetu loqi ehinh Dene6th~ehipnh~nlamh~th6ngTItaneilaBTL duQe.Tuynhien,h~th6ng phaneip eilaLLOCE ehig6m3 eip, DengiliacaelOpv~nkh6tlmm6iquail M vai nhau.Ngoaifa, so'luQngtu trongLLOCE eonkhahqnche(ehig6m 16.000ml,letu), Dennell mu6nap dl,lngvao mQth~th6ngth1,1'et , dn phai rnarQngthem. K~tluan:Dendungh~th6ngTItanLLOCE vaimQtso'di liensan: . MarQngv6ntu (d1,1'atIeDLDOCE vaWordNet) . Phancip sanhCln(d1,1'atIeDWordNet) . Themcaeki€u quailh~khaengoaiquailh~phaneip (d1,1'atIeDWeIdNer) . Hlnhthanhcaenhan nglinghlaeCiband~sudl,lngkhidn thietvacaeTItan naydn duQegQinha(gi6ngnhuCoreLex). 108 4.3.1.3 H~THONG NHAN NGU NGHIA TRONG MO IDNH BTL Quavi~ckhaosatcach~thongTItanngunghlad phh tren,chungtoi da quye'tdint sil'dl:111gh~thongTItannglinghlacuatiXditinLLOCE cungvoi mQt s6d.i tie'ntrenno d6 lamh~thongTItanngunghIachinhthuctrongmo hint BTL. Chungtoi chQnLLOCE lamh~thongTItanngunghlaVIh~thongnayco s6TItanvila phai (2441TItan),vila du d6 khil'nh~pnh~ngh~uhe'tcac tu c1a nghlathongthuong.Ngoaira,tudi6nLLOCE conchuanhi~uthongtinv~ngu phapkhac,dn thie'trhovi~ckhil'nh~pnh~ngnghlatvc1Qng.CachphanchialOp rhealoai,chungla~i,...cuaLLOCE rungphuh<;jpvoituduycilachungfa. LLOCE (xinxemth~mphVlvc 8.1)la tu di6nlOpnglinghlatie'ngAnh, nghlala khi tratrongtu di6nnay,ta se xac dint c1uQctrongm6i lOpc6 chua nhlingtu tie'ngJ\nh nao,nhun~chungta khongtrangu<;jcl~ic1u<;jc,nghIala tra tlr' naGthuQclOpnao.Trong lu~nan,chungtoi xem tu ai6nLLOCE la tu ai6n CED (Class-to-EnglishDictionary)gOGo£)6 tra ngu<;jcl~i (tiX"tlr'tie'ngAnh" suy ra lOp),chungtoi aa dva vao chinhtiXai6nCED nay (c6 ngu6ngoctULLOCE) c16 xaydvngmQttu ai6nngu<;jcl~i,co tenla ECD (English-to-ClassDictionary). Vin a~xaydvngcactuai6nphanlOpngunghIaaaau<;jctrlnhbaytrong4.1.3. *Vi dvmQtvai lOp(class)trongso2441lOpcuaCED: 0 LopAl (V): exist,be,create,animate,...(t6nt~ivat~osvt6nt~i) 0 Lop A2 (V): live, live on,exist,die,decay,decompose,survive,...(song/che't) 0 Lop AlSO (N): apple,apricot,peach,pineapple,pear,plum,papaya,cherry, grape,mango,dates,fig,pomegranate,...(traicay) 0 Lop G148 (N): letter,character,capital (letter),lower-case(letter), ... (chu cai) 0 Lop G155(N): letter,epistke,note,envelop,label,...(thutiX,ghi chu) 109 * Vi dlfmQtvaimlfctu(entry)trongso16.000mlfctucilaECD: 0 apple:AlSO a apricot:AlSO a exist:AI, A2,N1 a letter:G148,GlSS a bank:1104(money~"nganhang"),L99(natural~"basong"),..., Tuy Dillen,h~itudi€n CED vaECD trendaychili!.sl,l'phanlOpngfi'nghia ehocactu tiengAnh,d€ coh~thongphanlOpngfi'ngbiatUvl,l'ngtu'ongtl,l'nhu' - lIennhu'ngdanhchocactUtiengVi~t,chungtoi d1ixaydl,l'ngthem2 tu di€n phanlOpmoi, do la: tu di€n CVD (Class-to-VietnameseDictionary)va VCD (Vietnamese-to-ClassDictionary).Tli di€n CVD dungd€ traxemmQtlOpco nhfi'ngtutiengVi~tnao,contu di€n VCD d€ traxemillQttli tiengVi~tnaodo sethuQclOpnacoVi~cxayd1,1'ng2 tudi€n CVD vaVCD du'Qcdl,l'atrenb~lDdich tiengVi~tciiatudi€n LLOCE. Tieuchill,l'achQndatilamlfctUtiengVi~td~co th€ du'avaoCalltrucvi mo cuatu di€n chuyell du'Qcdl,l'arheacongtrinh[14]. *Vi dlfmQtvai lOptrongso2441lOpcuaCVD: . Lop Al (V): t5n t<;1i,t<;10ra, t<;10sl,l'song, ...(t5n t<;1iva t<;1Osl,l't5n t<;1i) . Lop A2 (V): song,tiep tlfc song,t5n t<;1i,chet,bu',thai, rfi'a,thai rfi'a,... (song/chet) . LopAlSO (N): lao,mo,dao,thorn,dua,Ie, m~n,dudu,anhdao,nha,xoai, eM la, va, ll,l'u,...(trai cay) . Lop G148(N): chfi'cai, m~utl,l',ky tl,l',chfi'boa,chfi'thuang,...(chfi'cai) . LopG155(N): thu',thu'daivaquailtrQng,thu'ng~n,~~n-gill chep,phangbl, baathu,nh1in,...(thu'tu,ghichu) *Vi dlfmQtvaimlfctu(entry)trongso16.000mlfctucilaVCD: 110 . ehfi'cai: G148 . metA136 (diy/la IDa),AlSO (tnEIDa),B82'(n~mmo) . t6nt~i:AI, A2, N1 (t6nt~i) . thu':G155(thu'tit),L177(hoanl~i) Caedi tie'nLLOCE nhu':Dangdung1u'<JngtU,phanea'pchitie'thondu'<Je chungtoithlfehi~ndlfatrenh~th6ngWordNet.Caehthuethlfehi~ndu'<Jctrlnh, bay!rangph~n4.1.3.Ngoaifa,d~d~theed6icaeke'tquagallnhanngil'nghIa, chungWidasud\lllgcaeTItanngil'nghIacobancilaCoreLextrong Bang8.8.Caemuephanea'pcobannhu',Hinh4.12sau: prt~I~eel - - pho:r I /T mle hum anm pit art fiat SPc~I~tme reI atr mea 41 ! lme qud qUI comtme ii fad ehm /\ grb grs )h\ Hinh4.12.Caemuephanea'pngfi'nghIacobantrongBTL. con pro 111 4.3.2 CACNGu6NTRITHDcD~XGLtNGDNGH~ D~ xli'19ngunghIa, ngu'aita phai ket h<;ipnhi~ungu6ntri thue:tU'cae tri thuev~ngonngu(nhu':hmhthai,nguphap,ngunghla)ehodencaetri thue ngoaingonngu(tri thuev~the'gioi th1,1'e)[136].Caengu6ntri thued6 thu'ang baag6m: 4.3.2.1 TRI THUC VB TV LOAI (POS) Trangtru'angh<;ipcaetu d6ngt1,1'(homograph)va e6nghlakhaenhauvoi caetulo~ikhaenhauvaungvoi mQttUlo~ichIe6mQtnghladuynha"t,thlnha thongtintulo~i,chungtasexaedintdu'<;ieehinhxaenghlaeuachung.Vi du:tu , "can"c6 nghla1a"e6 th~"(tr<;idQngtu), "cai hQP"(danhtu), "d6nghQP" (oQng . tu).VI V?y, voi cac tru'ongh<;ipnay, ne'ubier du'<;icchinhxaetu loai, chungta. heaDroankhli'du'<;ienh?pnh~~gnghlaeuachung.Vi d1,1:"IpRocanAUXCallyaDET eanNN"(Toi co ad donghl)pmQtcai hl)p).Vi~exacdinhtu lo~idadu'<;ietrlnh bay0ph~n4.2.1. TheothO'ngke trongtUdi~nLDOCE, e6 toi 88%m1,1etu thuQcd~ngn6i tren,ngoairac67%tru'ongh<;ipmam1,1ctu(t?Pcaetud6ngt1,1')e6nhi~utu lo~i, m6itu lo~ic6 th~e6nhi~unghlakhaenhau,nhu'ngtrangd6e6it nha"tInQttu lo~ic6duynha"tmQtnghla.DO'ivdi tru'ongh<;ipnay,tac6th~khli'nMp nhhg nghIane'utulo~icuan6(trongngucant) chinh1atulo~imachIc6mQtnghTa. Voi caeconsO'thO'ngke nhu'tren,takhongth~n6ilachIdungthongtintu lo?i,tac6th~khli'nh?pnh~ngnghIacuatutie'ngAnhvdidQchinhxactren88% vI2nguyenTItansail: . ConsO'88%la thO'ngke trenlo<:titu(wordtype)chukbongphai1u'<;ittU(word token),vatrenth1,1'cte,caetukhongroiVaGlo~itrenl?i xua"thi~nnhi~u. . DQchinhxaeeuabQgallTItantU'lo?ikhongphai100%(eB96-97%). 112 4.3.2.2 TRI THUC VE QUANH~CDPHAp (S-V-O-M)vA RANG BUQC NGUNGHIA(SELECTIONALRESTRICTIONS) Doi vdi caetru'onghQpclingtir lo<:iinhu'ngco nhi~ubonmQtnghia,th1 thongtin tir lo<:iikhongdud6khii'nh~pnh~ngnghIa.Vi d1,l:tir "bank"(co2 fir lo?i: dQngtir va daubtir)vdi tir lo?i daubtir co caenghla:"nganhang","bo (song)","day",.. Trongtru'onghQpnay,tadn sad\lngthemtrithuev€ the'giOi thljethongquacaerangbuQengil'nghIagiil'acaethanhph~neuphaptrangeau. Vi d\l: trongdiu "I enteranoldbank",saukhi quaph~ngallnhanngil'phap (4.2),tadu'Qe: [IpRo]NP[enterv[anDEToldADJbankN]NP]VPva cay eu ph3.pnhu' HInh 4.13 du'di day: ~.~ r .~ P r o verb ~ IP~ del adj noun I I I an °fd A-N ~ank I DoN If J 1 rLt &Y HInh4.13.Caequailh~euphapvarangbuQengil'nghIa. Trencayeuphapnay,taxaedintdu'Qecaequailh~ellphdpnhu:S-V (chu ngil'- dQngtir), V-0 (dQngtir- doitir),A-N (tinhtir- daubfir),D-N (dinhtir- daubtir).M6i tirthlje(contentwords)trangcauireD,ehodlidaxaedint du'Qcfir lo?i ehinhxae,nhu'ngd€u v~ngaynMp nh~ngv€ ngil'nghIa.Vi d\l: dQngtir "enter"(di vao/nh?p),daubtir "bank"(nganhang/bosong/day),tint fir "old" (gia/cG).VI V?y,chungtaph,hsii'dl:111gde'nnhil'ngrangbuQcngil'nghIanhu'sau: 113 Bang4.1.DanhsachcacnghlavarangbuQccuacactuth11ctrongcau. ~v~ y/ nt r rl \ ~ er2S- 'S- -0V-O Human Clo-SPA Human Data I- ~ I I T . T I bankl I lA-N old2 Hinh4.14.Cayquyetdtnhtrongvi~cchQnnghlaphilhc;iP. Quayi~cduy~tdiy tutrenxuongyei gocla dQngtu(Enter),cu6iclingta chQndu'~ccac nghlaphil hc;ip:enterl (di vao), bankl (nganhang)va old2 (cil). Trangvi~cxetdi~uki~nrangbuQcv~ngunghla,chungtaphaixetdentinhcap b~c(hierarchical)trongh~thongnhanngunghla(ontology)ma trongdo khai ni~mconsekethuacacnetnghlacuakhaini~mchavacothemconetnghla moiriengcuachung.Thongtinv~d~cdi€mngunghla(type)cuatUngtuth11c clingnhu'cac rang buQcda du'c;icxac dtnh trong tu di€n LDOCE va FrameNet[58]. Tu I RangbuQc/nhanngITnghla RangbuQc I (Toi) Type:Person(Ngu'oi) Enterl (divao) S:Human(ngu'oi) 0: Closed-SPA(khongkin) Enter2(nhp) S:Human(ngu'oi) 0: Data(dITliu) Bankl (nganhang) Type:Hou(nhacli'a,khonggiankin) Bank2(bosong) Type:Nat(congtrmhthiennhien,khonggianho) Oldl (gia) N: Ani (cos11song) 0ld2(cil) 114 4.3.2.3 TRI THUCVE NGONTV (COLLOCATION) Slf rangbuQcv~ngunghIagiuacacthanhph~nCllphapn6i trenkhong phaihienaoclinggiaiquyetdu'9CmQinh~pnhAng,vi c6nhlingquailh~ti~m5n v€ logic,v~ngu nghIaho~cth~mchi do th6i quell ma vi~cnh~nbier phai doi hoinhungtri thuc thegioi thlfCma dennay ngu'oita clingchu'ath€ richh9P her ,vaotudi€n haycaccdsatrithuckhactrangmaytinh. Vi d\l: danhtU "bank" trongcali "I go to the bank..."c6 nghIagi ? "ngan hang/ bo (song)/ day";danhtu "way" la "duong(di)/cach(thuc)";danhIU "letter"la "bucthu/ chucai"; ...NeutachIxetcacrangbuQcv~ngunghIanhu' phh IreD(khongphaihicnaocaerangbuQenay cli~~e6m~tday du) thila clingkh6maxaedintduQcchinhxacnghlacuacaetunMpnh~ngd6. Vi v~y,d€ khli'rrh~pnh~pglrangcaetru'onghQpnay,nguoitathuongxet Mn hint thai va ngu nghIaeua cae tu Ian e~nhay con gQi la ngon tu (collocation).Ch~ng h<;tnkhi thay "bank ... river" -7 "bo song", "bank ... account/money"-7 "nganhang"; "way to" -7 "duong(di)", "way of' -7 "each thue";"write ... letter...to" -7 "buc thu", ".. letter A" -7 "chu cai", ".. letters, digits,symbols..."-7 "chITeai", "write ...papers, letters,messages,... " -7 "b((c thu"';... Thongtin ve eactu c6 quailh~ngu nghlanhutren(thuQetruonglien Wang)e6th€ timtha'ytrangcaetudi€nthesauruseuaRogetho~eLLOCE. PhC;lmvi Iane~neuatudn khungunghlae6th€ labelltrai1,2hayn tuva benphai1,2hayntu.Ch~ngh<;tn,DavidYarawsky[150]diisud\lngeuas6ngu canh(context-window)c6 dQrQng[-50,+50],con Mark Stevensonva Yorick Wilks[136]sud\lng10tU Ian c~ntu dn khu nh~pnhAng,bao g6m:tu tbu nhat/thunhi bell trailben phai, danhtUldQngtu/tinh tu d~ulien bell trai/ben phai. 115 4.3.2.4 TRI TRUCvE CHl]BE (SUBJECT) Trong illQt so truongh<;5pnMp nh~ng,chungta c6 th~xac dtnhdu<;5cnghla dungcua tll DeUta bier dU<;5cchu d~cua van b,ln. Ch~ngh~ntll "bank", neu dangn6iv~vand~v~lInt vlfc"taichanh"thlnothuongnghlala "nganhang"; tll'"driver"~ "trlnhdi~ukhi~n"(neuchud~la lInhvlfc"tinh9C"); "sentence" -7 "cau"(neuchud~la "ngonngu/ vanph(;lm")ho~c"banan"(neudangn6i v~"phap lu?t"); "element"~ "nguyen to'" (trong"hoa") / "ph~ntU'" (trong "toan/tinh9C"), ... B~ xac dinh dudc chu d~cua van ban dan£dn dich, ta dn xem xet su. . ~. . xuathit$nmQts2 tll chuyenmantronglInhvlfcd6.Ch~ngh~n,neutrongvan bantathayxuathi~ncactll nhu:"ellipsis"(tInhlu<;5c),"bilingual"(songTIger), "anaphora"(the d~iill), "phr~se"(ngu),...th1ta c6 th~GOaDnh?n van ban nay dangn6iv~chude"ngonnguh9C";tudngtlfchocactll "computer",memory", "peripherals", "CPU",...~ dangn6i ve "tin hQc";... ChinhVI V?y, trong tU di~nLDOCEILLOCE d~uc6 ma s6 chu d~cho cac tll ehuyenmannay.Chungtac6th~xacdtnhdU<;5cchud~mQtcachtlfdQngb~ng eachxemxetcactll chuyenmanIanc?n tll dangdn khU'nMp nh~ng(gia sti trongph~mvi [-50,+50])thuQcchud~naotheocongthuccuaYarowskytrong ~ Pr(wlSCa0Pr(SCa0 [150]nhusau:ARGMAJ(;colL.,log P 'C.) trongd6SCat:machu'1'EW 1 VI, d~,W: khungcU'as6ngucanhchuacactll w.Do xacsuatPr(SCat) khongph1;1 thuQc vao w, Den cong thuc tIeD du<;5c viet thanh: ARGMAX" 100 PrCw ISCat)SCaI,L.; b WEW PrC w) (4.13) Tacoth<ith6ngkecaccons6trentIeDmQtnguli~ucionngu(tiengAnh) 16n. 116 4.3.2.5 TRITHUCVE TAN SUATNGHIACUATV (SENSEFREQUENCY) Khong phai tu naG rung thuQcv€ mQt chu d€ naG d6 (trong tU di€n LDOCE,hon 56%tu thuQcd;:tngnay), VI V?y tinhthongdt,mgcuamQtnghlanaG d6condu'<;fCd1,1'atrendQdov€ t~nsua'txua'thi~ncua tu d6voi nghla C1,lth€ d6. Chhg h;:tn,danhtu "pen" se c6 nghlathongd1,lngnha't13."but/vie'C'(ben qnh caenerhlait thonerdUllerhon nhu""chu6ner""lonerchim")' "ball" c6 thu'onerc60 0.0" 0' 0' 0 nghlala "quabanh/honbi" honla "bu6ikhieuvu",... BQ do t~nsua'txua'thi~ncuam6inghlacuam6i tu du'Qcthongke tren nhungnguli~ura'tIOnthuQcnhi€u lo;:tivanbankhacnhau.ChinhY1v~y,trang WordNetva trongLDOCE, 'cacnghladu'Qcs~pxe'prheathli't1,1'giarnd~n(nghTa thongd1,lngnha'tseduQcli~tked~ulien).. 4.3.2.6 TRI THUC TRONGDJNH NGHIA CUA MOl ~GHIA(DEFINITION) Trongcactu di€n LDOCE / WordNet,m6inghlaseduQcdtnhnghiavavi dt,lkernrhea.Vi d1,l:tu "bank"trangLDOCE se c6 cacnghlakernd~nhnghla cilanonhu: -"landalongthesideofariver,lake,etc."(da'tdQcbellsong/ h6) -"aplacewheremoneyiskeptandpaid " (noigillti€n vatrati~n...) - "arow,a lineof..."(rnQthang,mQtday...) D1,1'atrenthongtin trongcacdinhnghlanay,va sosanhvoi thongtin cua ngucanh,tac6 th€ xacdinhduQcnghlaphuhQpcuatutrongngucanhdo.D€ th\fChi~ndi~unay,Wilkset.al.[147]datinhtoaDph~ngiao(overlap)cuata'tca caet6h<;fpnghlacuacaetu th1,1'Ctrongcali tie'ngAnh dungd€ dinhnghlam6i nghlacuatu. 117 4.3.3 CAN NIlAN NGU NGRIA CRO TV TIENG ANR Voi h~th6ngnhanngunghladuQctrlnhbayd tren,th1bai roankhli'nMp nh~n.gn hlacilatll duQcduav~bai roangallnhanngunghla.Tucla nghlacila tITdanghlaseduQcxacdinhngayne'ubie'tnhanngunghlacilano,vi d\l:danh tIT"bank"se co nghlala "nganhang"ne'uduQcgall nhanla "HOD", va co nghlala "bo(song)"ne'u,duQcgallnhanla "NAT", Wongt1!rhocaetll "letter", "line",.... Trongcacmo h1nhgallnhanngunghlarheaeachtie'pc~nd1/atrencac ngu6ntri thuGnoitren,nguoitathuongsli'd\lngbQnhancodQmill (granularity) khaenhau.BQ nhanrang min (chi tie'thangtramngannhannhuWordNet) thl ~ oQehinhxac cila vi~cgall nhan se tha'pbon nhungkha Hangkhli'nh~pnh~ng nghlacilanosecaobon(vIkh_ongcotruonghQpnaGclingnhanmakhacnghla). Ngu'Qel<;li,ne'uchQnbQnhancangtho(chico36nhannhuLDOCE),thldQchinh xacsecaobonva tatnhienkbaHangkhli'nh?pnh~ngnghlasethapbon(seco nhi~utruonghQpclingnhannhungkhacnghla). Ngoai fa, vi~cgall nhanngu nghla con duQcphanbi~ttheequy mo gall nhan:ho~cla gall chomQtso'it cactll di~nhlnh(nhu HweeNg va Hian Lee [89]eho1ill interest,David Yarowksy [151]cho 12tll, ...)ho?c la gall cho hgu h€t cac tll th1/c(nhu Mark Stevensonva Yorick Wilks [136],Mona Diab va PhilipResnik[66]). Vi~cchQnngu6ntri thuGnaGrhomoitlnhhu6ngduQch~th6ngquye'tdint b~ngphuongphaphQcgiamsatlIen nguli~udi'iduQcgannhanngunghlachinh xac(daychinhla nguli~uhuanluy~nhaycongQila nguli~uvang).Giai thu~t hQcoth~la m<;lngNeural,cayquye'tdint, MBL, TBL,...matrongdo cacgiai thu~th9Cd1/atrenkyhi~u(symbolic)torachinhxacbon[112]. 118 Trongvi~chQcgiamsat d€ chQnngunghlathichhQpnhuda trlnhb3.y d tIeD,thlnguli~uhuanluy~nla vanderatdangduQcquailtamVInguonnguli~u nayrat hiem(chi c6 bQSEMCOR chilakhoclng250.000tU trichtu ngu li~u BrownvaduQcgallrheanhanWeIdNer)vavi~cxaydl;fngn6rattonkern.Chinh VII:;'do nay, khien cho caebQkhli'nh~pnh~ngngu nghlahi~nnay chu yeti chi khli'cho mQtso it cac tu, VI neu muonkhli'cho tat eelcac tu (hang Vt;llltu da. nghla,m6ituphclixuathi~nvaitraml~n),thlnguli~uhuanluy~nphaichilavai tri~ututrdlen.Neusonhandlllg IOn(nhuWordNet),thlnguli~unaycangphai IOnhon(hangtytu). D€ kh~cpht,lckh6khanve nguli~uhuanluy~n,mQtsotacgia trongcac eorigtrinhnhu[151J,[66J,[62Jdachuy€nsangdungnguli~ukhongdanhdati VahQcb~ngphuongphaphQckhonggiam sat,nhungtatnhienket qua se thap hoDphuongphaphQcgiamsat. MQtxuhuangg~ndaynhatla dungnguli~usongngu(mada duQcdiet ehinhxac bdi nguei) d€ xiiy dt,l"ngngG'li~uhuan luy~nehoWSD. Y tudngnay duQeBrownvacaedonguicghidadexuatd~utieDtrongcongtrlnh[51].Theo ong: "M(Jt tli do nghfotrongngonngilngu6nthuangdu(Jcdjchthanhnhilngtli kluicnhoutrongngonngilarch...Diiu naycangdungclIocacc{jpng6nngilkhac lo(lihrnh".Vi dt,l:tU"bank"c6nhieunghla,nhungtrongmQtngil'canhcuamQt bandichchinhxacbdinguei,n6 seduQcdichthanhho~cla "nganhang"ho~c Iii "besong",...vadaychinhla "nhanngG'nghladung"cuatU "bank"trongngil' li~utiengAnh.MQttrdng<;ticuad€ xua'tnayla sl;fkh6khantrongvi~cxiiydl;fng ngG'li~usongngu1-1d6.Nnungg~nday(2000-2002),khinhungnguonngil'li~u songngil'nhuv~yduQchint thanh,thl tIeDthe'giai da c6 nhi€u cong trinhchQn eachtie'pc~nnay,nhu:[66J,[101J,[90J,... 119 14.3.3.1 GAN NHA.NNGU NGHIA CUA TO'TRONGMO HINH BTL Trangmohint BTL, chungtoi dach9nh€; th6ngTItanLLOCE vdi mQtso' 5sungnhl1datrlnhbayaph~n4.3.1.3.Vdi h€; th6ngnhanngii'nghTanay,biii DanWSD trangmohinhBTL dl1Qcdl1av~bai tmingallTItanngunghTa.Th?t ~y,nghlaeuatu "letter"se dl1Qcxac dtnhngaynEubiErnhanngii'nghTacua 6:nEuno dl1Qcgall nhanla G155,thi no co nghI~la tInt,G148-+ chll c6i; u'clngn;reho tu "bank", "way", "line", Do h€; th6ngTItancuaBTL chI g6m Gang2500TItan,Denvdingii'li€;ukhoang5.000.000tu,thisl.l'pMintandu1i~u 6ngxayra. TrongmohinhBTL, bai roangallnhanduQcthl.l'Chi€;ntrenta'td cactu hl!c(contentwords)baag6m:danhtu,dQngtu,tinhtu,d,;iitu.D6i vdicaetuhu functionwords),chllngtoi sesii'd1,lngcacthongtinv~ngG'dnh d~chQnnghia hUhQP.Ch~nghq.n:gidi tu "in" seduQcdiet la "trang"nEu danhtu theesail 6chiv~khonggian(nhanla SPA); se duQcdich1a"b~ng"nEudanhtu do huQcv€ vanban,ngonngii',..(TXT, LING);... Trangmo hint BTL cl'1achungtoi, chungtoi da kh~cph1,lctra ngq.iv~ngG' i~uhua'n1uy€;nb~ngcachtl.l'xay dl.l'ngngii'li€;u hua'nluy€;n dl.l'atren ngii'li~u ongngG'da duQclien kEt tu [67].Vi€;c tl.l'xay dl.l'ngnhuV?y se g~p nhi~ukho an(v~ngu6nngii'li€;u songngii'dadl1Qcdichchinhxacbaingl1oi)nhungding eml<;iikhongit thu?nlQi (lamchuduQcngii'li€;u, ngu li~uphuhQpvdi lInh v1,1'C h dich).Chungtoi may m~nkE thuadl1QcmQtngu6nngii'li~usongngu Anh- i~tdi~nHi cotenlaEVe tueongtrlnh[13].Eve cokichthudc5.000.000tu, huyenv~khoah9Cvakythu?t(Tinh9C,Vi€n thongLYi~clienkEttuehangu i~uEVe nayda duQctrlnhbaychi tiErtrongph~n4.1.1.Vi~cgall TItanngii' ghiachongii'li~uhua'nluy~nseduQctrlnhbaytrangph~ndudiday. 120 4.3.3.2 GAN NHA.NNGG NGHIA CHONGG LI$U HUAN LUYJ;,N EVC Ngfi'li~usongngfi'EVC g6mkho?mg250.000c~pCalinhuduoiday: (E):Jetplanesfly aboutninemileshigh. (V): Cacmaybayphanh,tcbaycaDkhoangchind~m. Giasaxettu"plane"(g6mcacdanhtu:"maybay","matph~ng";dQngtu: "bay,lu<;1n","di maybay","bao";tinhtu:"b~ngph~ng")va "fly" (dQngtu:. "bay","tha", danhtu: "conru6i",...) trongcali tie'ngAnh n6i freD,chungd~ula nhungtu da tu lO<;liva danghla,nhungsailkhi xac dinhdu<;1cmoi lien kef tu chinhxac(ph~n4.1.1)vatUlO<;lichinhxac(ph~n4.2.1)taduQc: [JetNN planesNNP]NP[flYvJvp ~ \ [aboutRB milesNNP highRB]MP [Caemaybayphanl1,rc]NP[bay]vp [caD khoang chin d~m]MP Trangd6 : "plane"la danhtu,"fly" la dQngtu; "plane"lienketvoi "may bay","fly" lienketvoi "bay".Duatrenthongtinnay,chungtoi lien hanhxac dinhnhanngfi'nghlaciladanhtu"plane"vadQngtu "fly"nhusail: GQiejla tu tiengAnhthuQcCalitiengAnhe,vi la tutiengVi~tthuQccali tiengVi~t v ma dU<;1clien ket voi ej.Do eiva \j d~uc6 th~la nhfi'ngtll'da nghla, Denejva vi d~uc6 th~thuQcnhi~ulOpngfi'nghlakhacnhau(nhungv~nclingtll' 1o?i).GQi Xi 1amQttrongnhfi'nglOp ngfi'nghla X child eiva Yj la mQttrong nhfi'nglOpngfi'nghIaY childvi'VI ejvavi la tu du<;1cd~chboinhau,DenlOpXi phaitrlingho~cg~nnghlavoi lOpYj nha't.VI v~y,nhanngfi'nghladungcilaej va. vi chinh 1a lOp Xi va yJ du<;1C xac dinh nhu sail: ARGMAX X;Ex'yjEyClassSim(Xi' Yj) (4.14). trongdo,ClassSim(Xj,Y)du~ctinhrheaGongthuG(4.3). 121 Vi du:xetdanhtu "plane",1atha'y"plane"thuQcaclOpX={J41,M180}, "maybay" thuQccac lOp Y={M180},Den khi tinh e~p(X,Y) ta duc;>c (M180,M180).V~y nhan ngil'nghIadungcila "plane" trongTruonghc;>pnay 1a M180(phuongti~nhangkhong).TuongtVehodQngtu "fly" va "bay", tae6 X={M19,M28,M29,M30,M35}va Y={M19,M28,M35}.TrangTruonghQpnay, tae6th~e6tdi 3nhand<;1tcved<;1irhea(4.14),d6la: M19,M28vaM35,nhu'ng, dotinhtoaDdQtuongd6ng(trangGongthue4.3),tasechIduQcM191anhanciia "fly".Chitie'tv~vi~cxaydVngnguli~uhua'nluy~ndaduQcGongb6Trang[67]. DVatrenGongthue(4.14),chungtaclingd6ngthaixaedinhduQcnhanngu nghIachinhxaccilatu tie'ngVi~tTrangnguli~usongngiIEVe. Ngu1i~utie'ng k Vi~tduQcgall nhiinngunghlanayse giupfchra'tnhi~uTrangvi~cxaedtnh (b~ngcachth6ngke) pho.ngeach,ea'utruc,tr~ttVtu Trangcautie'ngVi~t,g6p phht<;1oranhungcautie-figVi~tggnvdivanphong,ca'utructVnhiencilanguoi Vi~tbon.Ngaaira,ngu6nngerli~utie'ngVi~tduQcdanhda'unay se13ngu6n nguli~uva rungquigiadoivdi caenhangonngerhQesosanhAnh-Vi~tTrang vi~csosanh,d6i chie'ucaedi~mwongd6ngva dtbi~tacac ca'pdQ IreD cae blnhdi~nkhac nhau.Chinh caenghienCUDnay, l<;1id6ngg6p nguQcfro l<;1irho vi~cehQncacye'uto'lac dQngTrangvi~cehuy~~nger(th~hi~nTrangcac khung lu~teilagiaithu~thQcKFTBL) tuAnhsangVi~tTrangmohlnhBTL. 4.3.3.3 GLA.ITHUAT HQCBE GANNHA.NNGD'NGHlA TU TRaNGBTL Giai thu~thQcduQcSUd\lllgTrangmohlnhBTL 13giaithu~thQcgiamsat KFTBL (diiTrinhbaychitie'tTrangphgn3.3.4.6)vdih~th6ngnhanLLOCE di tie'nvangerli~uhua'nluy~nEVC diiduQcgannhannguphapvangunghIa. 122 GannhanngunghlachotutiengAnhclingHiIDQtd~ngdi~nhlnhcuavi~c gallnhanng6nngubiingTBL. Sovdicacbailoangallnhanng6nngukhac,bai lOannayyetidu d~uVaGla nguli~udaduQcgallnhantulo~ivaell phapehinh xac. Gonhanngunghla Nguli~uhuinluy~nEY dagallnhantUlo~i,Cll phapvanglinghIa Nguli~uchiconhan nguphap Cacngu6n tri thlicv~I ngunghla trongtudi~n BQ gallnhan ngunghlacd sa ,"""""""""""""""".. .. . , ~~~~.~l.~~t..j Ngu li~uduQcgall nhanhi~nhanh ,. Caclu?t ling VieD Ngu li~ugall nhan rhealu?t ling vieD Lu? t toi U'U Hlnh4.15.M6 hlnhhuin luy~nbQgallnhanngunghIatutiengAnhtrongBTL. 123 Daylu~tdi bie'nduQcrutraseduQcdungd6 gallnhanngunghIachQtu tie'ngAnhtrongvanbanmditrongh~dtchEVT. Van bannaydn philhQpvdi vanphongva llnhvl1ccuangfi'li~uhua'nluy~nEYC. Bai roang5mcacva'nd~ sail: Gannhancdsd(baseline): Vi~cgallnhancd sa duQCthl1chi~novatIeDthongtin v~t~nsua'txua't hi~ncuanhanc doi vdi nhfi'ngtu w mangtu lo<;iip trongngfi'li~uhua'nluy~n EVC:P(clwp)'Tu lo<;iip la tulo?i chinhxaccuatuw duQcxacdinhsailkhi qua thg gall nhantu lo<;iicua BTL. Boi vdi tu mdi (khongco trongngfi'li~uhua'n luy~n),thongtin v~dn sua'tnhanngfi'nghIanay ducjcla'ytu ngu6ntri thuc \: WordNet(ta'tnhienla cotinhde'nthongtintu lo?i vachUd~).Cachgallnhancd sanhutIeDd1'ith~hi~nvi~csli'dt:mgngu6ntri thucv~tulo?i,tri thucv~chud~ vatrithucv~t~nsuatngfi'nghIamadlltrlnhbaychitie'trongph~n4.3.2. Khungluat(template): Caekhunglu~trongbairoannaysesli'dt,mgcacnhantotucacngu6nIri thuccolienquailde'nvi~ckhanMpnhangngfi'nghIa(ngoainhfi'ngngu6ntri thucda:duQCsadt,mgtrongph~ngallnh1'incdsaa tIeD),c\lth~nhusail: . sa dt,mgtrithucv~quailh~cuphapvarangbuQcngfi'nghIa: n((Syn, E TagJ 1\Word; 1\(Wordj E Tag}) 1\POS;) ~ (Wordk / Synk +- Tag k) vdi 3;e[-m,+n] Synla bie'nth6hi~nvaicuphap{Sub,Obj,Noun,Adj,Verb,...};Tagla bie'n th6hi~nnhanngfi'nghIa{HUM,ANI, NAT, M1,...};Wordla bie'nth6hi~n hlnhthaitu(nhu:"of", "account",...);P~Sla bie'nth~hi~ntu 10(;li{NN,VB, DT, IN,..}. . sa d\lngtrithucv~ngontu(collocations): nWord;, nWordj I\POSj hay nTag; ~(Wordk/Synk +-Tagk) 3;e[-III,+II] 3;e[-.'.+IIJ 3ie[-III,+II] 124 Day luatt6iu'u: Sailkhihuanluy~n,h~th6ngnayserUtraduQcdaylu~t6iullcod<:tnghusail: 1. ((SUBoE HUM) 1\(Wardo="enter')1\(POSo =VB)1\(WardoEMOV)~ OBJo *- HOU co nghlala: CHU TV (duQcgallnhanngunghlala NG001) cuaBONG TV A? -t', "enter"vdinhannghlala Dl CHUYEN thiBOl TO cuanoseduQcgallnhan ngunghlala ToA NHA. Vi dl,l:"1enterthe bank"thld6itu "bank"sedu'Qc gallnhanlaToA NHA (nghla"nganhang"). 2. ((Wardo="way")1\(Ward+i="0/")) =>toga*- MANN conghIala: tu "way"seduQcgallnhanngu~ghlala CACH THUC nelili~n sailnola tu"of". 3. ((3iE[+1,+3]IWort!;="river"JI\(Wordo ="bank)1\(POSo =JvW))=>tag0*-NAT co nghlala: danhtu "bank"se duQcgallnhanngunghIala CONG TRINH THIEN NHIEN (bosong)neusailnocotu"river". 4. (Wordo="in")1\ (POSo=IN) 1\ (NOUNoE LING)=> tag0 +-"bang" co nghlala: giditu "in" seduQcgallnghlala "bang"neusailnocodanhtu chinhcuagiOingu"inNP" dothuQcv~ngonngu.Vi dl,l:"inEnglish". 5. (Word-i E SPORT) 1\(Wordo="l71atch")1\(POSo =NN)=>tago*-SHOfiV co nghlala: danhtu "match"se duQcgallnhanngunghlala BUGI TRINH DIEN (nghIala "tr?ndau")neuli~ntrudcnola danhtu thuQcv~ THE THAO. Vi dl,l:"footballmatch". Lull v: dc nhanngunghladuQcgallnoi trenchom6ituphaiDamtrongdanh sachdc nhanngunghlaco th~cocliatu do (vditu lo<:tixacdtnhcuano).Ket quagallnhannayduQcdanhgiabangeachsosanhvdinguli~uyangnhuTBL (3.3.1.5). 125 ~ ~ A , , 4.4 CHUYEN DOl CAY CD PHAP TRONG BTL Trang h~diet theaehie'nluQeehuy~nd6i ell phap(syntaetie-transfer- based),thl vi~eehuy~nd6icayellphapdongvai trara'tquailtrQnglmhhuang de'neha't1uQngdiet euah~.Chuy~nd6icayell phapnh~IDehuy~nIDQtdiy ell phapcuang6nngungu6n(dayla tie'ngAnh) sangcayellphapcuang6nngu diehd~philhQpvoi vanph<;lIDeuang6nngudieh(tie'ngVi~t).Vi d\l:Calitie'ng Anh "] enteran oldbank"saukhidi quacaedng phantichhint thai,nguphap vangunghla,taduQe: Bang4.2:Thongtinphantichngonngucuacalitie'ngAnh. Dlja tIeDcaethongtin v~ngonngunay,ta tie'nhanhchuy~nd6icayell phaptutie'ngAnhsangtie'ngVi~tnhuHlnh4.16duoiday: /~ /~ r ~ r ~ del adi noun del noun adi I I I I I I enter an old bank T6i diviw mi}tnganhimgcu] Hinh4.16:Chuy~nd6i cayellphapngonngungu6n.sangcayngonngudich. Hinhthai' I I enter a old bank Tu la<;li I PRO V DET ADJ N 1 Cll phap I []NP - [ [ ]NP]VP Vai i Agent Act Qty Time Location I I HUM MQV INDEFNaunahla I TME HOU;:;, ;:;, Lien ke't ' 1 I 2 3 5 4 I! 126 , , A' A A A? A? A , , CAC CACH TIEP CAN CHO VI~C CHUYEN DOl CAY CD PHAP:.4.1 D~ giai quye'tbai loan nay,ngu'oita cHie6nhi€u eaehtie'pe?n khaenhau, ,hu':eaehtie'pc?n dl1atren Iu?t, quabi~udi~nbell trong,phan tIch ell pM p ongsong,thO'ngke tr~ttl1,hQctUnguli~u,..Saildaytaxemxet so Iu'<;1ccac aehtie'pc~nd6: ..4.1.1 CACHTIEPcAN DUNGLuAT (RULE-BASED) Day Ia eaehtie'pc?n du'<;1Capd1,lngsamnha't(tiInhungth?pnien 60-70) )~ngcachdungnhungIu?tehuy~ndcheO'dinh.doconngu'oifighTravag~nIi€n lingIu?tsinhcuabQvanph(,tmdungd~phanrichCallngu6n.Vi d1,l:d~chuy~n 16iVl tri giuatinhtiI~danhtiI trongtie'ngAnh voi tie'ngVi~t,ta dungIu?t :huy~nd6i sail [6]:"(E): 1\T --+Det Adj Noun =>(V): NP --+Det Noun Adj".. \TghTal khiphanrichellpMp calltie'ngAnh,ne'utrongcayCllpMp ngu6nc6 ;ud1,lngIu?tE, thlkhi chuy~nsangcayellphaptie'ngVi~t,cayd6sedu'<;1cvie't ;:titheoIu?t V tu'ongling.Voi eachanhx(,tnhu'tren,taclingc6 th~giai quye't :actru'ongh<;1p"chen"ho?c "xoa"nhungnodekhi chuy~ncay tiI ng6nngu 19u6nsangng6nnguc1ich.Tuynhien,eachthlicnaychIgiaiquye'tvi~cchuy~n 16i!chenixoacnocaethanhph~ntrongcungmQtIu?tsinh(nhu'RlsangR2 trong -nub4.17). NghTaIa n6 kh6ng th~chuy~nIDQtnodesangIDQtca'pkhac ho?c ,angIDQtchakhac (nhuR1 sangR3): R1 A~ /1B~ R2 A4\ /1C~ B2 Bl81 B2 C1 C2C1 C2 A4\ /1B Cl C ~ Bl C2B2 Hlnh4.17:KhaDangchuy~nd6i cayellpMpb~ngIu?1. 127 4.4.1.2 CACH TIEP CAN DUNGNGD'PHAp SONGSONG(ITG) Bay la cachtie'pc~ndvatrenvi~cphanrichcuphapsongsongchosong nguAnh-Vi~tcuaDekaiWu [148]chosongnguAnh-Hoa.Vi~cphanrichCll phapsongsongnay su d\lngbQvanph<;lmchuy6nd6i nguc;lc(ITG: Inversion TransducerGrammar).Van ph<;lmnay duc;lcbi6u di~n:G=(Nl,Wl,W2, R, S), trongd6: - N: t~phc;lPhUllh<;lncaeleihi~uchuake'tthuc(nonterminal), - WI: t~phC;phUllh<;lncacleYhi~uke'tthuc(terminal)cuangonnguthuI, - W2: t~phc;lphUllh<;lncactuvvng(terminal)cuangonnguthuII, - R: t~phullh?n cac lu~tsinh - S E N la leihi~ukhdid~u. Khi d6, khCfng ian cac C?PJu (terminal-pairs)X =(WIU {c:})x(W2u {c:})chua nhungC?Pdichtuvvng(lexicaltranslation)duc;lcth6hi~nbdix/y,connhungtu adndQc(singleton)duc;lcth6hi~nbdi x/£ hay £/y, trongd6 xEu;~va y EW2. M6i lu~tsinhse c6 dang:hoacla tr~tv blnhthuongA--+[aj,a2,...,arJ,hoacJa tr~tVblnhdao A-'7 vdi Q;EN u X va r la b~ccuacaclu~tsinh. T~phc;lpnhungchuy6ndtchduc;lcsinhbdiG auc;lckyhi~ula T(G). Chungtoi dil apd\lngvanph::tmnaychovi~cchuy6na6icaycuphapAnh-Vi~tdvatrensong nguAnh-Vi~tEVC (chitie'tail trlnhbaytrong[17]va [28])vdike'tquac6dang sail: [[[If/Ne'u] [a/mQt]NP [controls/ai~ukhi6n NP]VP ]SBAR [,I,] [you/b?n]NP[should/Den[aSk/yelldu [the/(administrator/nguoiqUailtrt]NP] VP]VP [to/( [install/caid?t [the/(<proper/aung d~n rand/vasoftware/ph~nm~m]J>]1\TpJVPJVP[./.J]. [hardware/ph~ncling .>..:..u 4.4.1.3 CACH TrEP C~NDUNG DJCH THONG KE (SMT) Cachtie'pc~ndungvanph<:lmsongsongITG n6itIeDv~ndlfavaobe)phan richcuphapdlfatIeDlu~t,Denke'tquakhangcaosovei caccachtie'pc~nmei hi~nnaydlfatIeDvi~chQctunglili~usongngli.Me)ttrongnhlingcachtie'pc~n meid6la thongke tIeDnglili~usongnglida:dU'Qclienke'ttu,d€ rutraxacsutt chuy€ndichvi tri (transposition)cuacactutrangngannglidichsoveinganngli ngu6n.Day chinhla eachtie'pc~ncuah~dichthongke SMT (da:gidi thi~ua 2.2.2).Theacachtie'pc~nnay,ne'utac6 Calitie'ngAnh e, dn dichsangcali tieng Vi~t v, thi vi~c nay tU'dng dU'dng vei vi~c di tim Cali v =argmax P(v) *Pee Iv). Trangd6,taphaitint dU'Qcxacsuttmahintngan,. ,, ,. ngli PCv) va xac sutt ma hint dich peeIv). Ta tint PCv) nh~md~tim cali dung ngliphap(dt nhienla phiiidu'ngv~tr~tlftu)tie'ngVi~t.D€ tinhxacsuatnay, chUngtaidi:isud\lllgcaerangthueduaiday[19]: P(V)=PCV1V2...V,)=PCVI)*P(V2IVj)*P(V3IvIV2)*"'*PCV, Ivlv2",v,-I) (4.15) Ne'uchungtagiasur~ng1£llakichcBcuattt ciinhlingtuvlfngtrongtudi€n tie'ngAnh,thisotrU'onghQpkhaenhautungdaimQtcuae1e2...ek-1axacsuttdi~u ki~nthuk sela 1£('.Day lamQtcansortt IOnvadod6chungtakhangth~tint mQteachchinhxacrhome)tsoh<:lngPCVkIVjV2...Vk-J)duQc.Vily dod6,chung t6isud\lllgmahint trigramd~tint xtp Xl rho tungsoh<:lngtrang(4.15)rhea eachsail: PCVk IVjV2",Vk-I);::O P(Vk IVk-2Vk-j) (4.16) C6nghlala thayvi xctk-1tuvlfngVIV2",Vk-ldaxayraa phiatruactuvlfng Vk thichungtaigiaih<:ln<:lichIxct2tuvlfngaphiatrU'acdi:ixayrathai.Moi be) ba(Vk-2Vk-lVk)dU'QcgQila trigram.NhU'v~y,tu (4.16)chungtac6 th~vie'tl<:linhu sail: 129 I P(V)=P(VjV2",Vk)=P(V1)*P(V2Iv1)I1P(V; IVI-2V;-1) 1=3 (4.17) Chungtathiy rangc6th6c6Id trigramklulcnhau,vac6nhungtruongh<;5p trigramkhongbaogiGxuit hi~ntrongnguli~uhuin Iuy~ncilachungta.Luc d6 se t5n t~i mQt P(Vk IVk-2Vk-1)=0,vadi~unaysed~nde'nP(v)=0.Vily dod6, chungt6i rnotaP(VkIVk-2Vk-I)b1ngcachtinhg~ndung(lamiron: smoothing)nhu sail: P(Vk IVk-2vk-j)=B1 *T(v* Ivk-2v*-I)+B2 *B(Vk !vk-j)+B3 *U(vk)+B4 (4.18) Voi: 8j, 82,83,8lIa cach~so(h~ngso)va81« 82« 83« 84 T: xacsuit trigram(batll'lientie'p) B: xacsuit bigram(haitll'lientie'p) U: x:icsuit unigram(d_uynhitmQtll')vab~ngl/lel ChungtoitinhGongthuG(4.18)lIennguli~utie'ngVi~ttrongsongnguAn11- Vi~tEVC d~xacdint tr~tVtliphilh<;5pnhit chotie'ngVi~t.Tuynhien,ke'tqua khongdU<;5cnhu mong d<;5i(chI khoang60%, kern hon so voi 2 each tie'pcan d~u).Di~unay chothiy ne'uChIxet lIen b~m~t nhueachtie'pc~nSMT nay, thl hi~uquasethip. Chinhvi v~ymatrongcaeGongtrlnhg~nday,nguoitadi'itint xacsuitP(v)naylIeneelciu truccaycuphap[149]. Tuy ke'tquaconh~nche',nhungr6 rangphuongphapnayc6 u'udi6mv~ phuongphaplu~nla tv dQnghoanroan,chungtakhongdn xay dvngt~plu~t chuy~nd6i band~unhu2 tie'pc~nd~u(lu~tco dint vavanph~rnsongsong). Honnua,c6nhungchuy6nd6ivi tri made'nnaychungtadIng chuat6ngquat hoathanhquilu~tdU<;5c(c6th6doth6iquellhaydo_mQinhantonaod6).Vi v~y, chungtoi v~nkhaithacke'tquaxacdint tr?ttv tli cilacachtie'pc?nnaynhumQt thamsotronghamquye'tdinttr?ttl,rtlicilacalitie'ngVi~t. 130 4.4.1.4 CACH TIEP CAN DUNG NGD LI~U (CORPUS-BASED) Cachtie'pc~nnaysedungGongngh~mayhQc,hQctrenkhongii'li~usong ngii'dadu<;1clienke't(tiI',ngii')vaphantkh cuphap,d€ 1\1dQrtgrutra cacqui lu~tchuy€nd6icaygiii'a2ngonngii'.Trangcungxuhuangnay,nguoitadathl!C , hi~nb~ngnhi~uphudngphapkhacnhau,di€n hinhla: phudngphapTAG d6ng bQ du<;1CgQi t~t la STAG (SynchronousTree-Adjoini,ngGrammar)[131]la phudngphapchuy€nd6i dl,l'atrencaccaysddip cuavanph~mTAG[92);thu~t. loan baa loan nut eha (dominance-preserving)va rang buQcLCA (Least CommonAncestor)[108];... Vai cach tie'pc~nhQctrennguli~uki€u nay,bQlu~tchuy€nd6i (tr~t l,l' tu,them/xoa,...)se du<;1c\1dQngrutra tu nguli~uthl,l'cte'.Ne'unguli~uhua'n luy~neangd§yau,baaquat,thlbQlu~trutracanghi~uquavachinhxae.Tuy nhien,d~codU<;1cmQtkhonguli~unhuv~yla di~ura'tkho,VIv~ykhongphai M thongnaGclingco th~sli'dl;lngcachtie'pc~nnay.Ngoaifa, co nhungsl,l' ehuy~nd6ib~tbuQedod~cdi~mloaihInh,nhungchuy~nd6inaynenc1u(jc1u'a thanhlu~tcodinh(corephenomena)vachIDend~h~thonghQccaelu~tehuy~n d6iphl;l(marginphenomena)nh~mtanghi~uquacuah~thong. 4.4.1.5 , "" A , ,:! ~ CACHTIEP CANDUNGBIEUDIENTRUNGGIAN ThayVIehuy~nd6i trl,l'ctie'pnhu4 cachtie'pc~;tren,nguoitaconsli'dl;lng sli'dl;lngki~uchuy~nd6i thongquamQtbi~udi~ntrunggian.Ch~ngh~nthong quabi~udi~n"khungnguphapcach"(caseframe)[144]cuacall(lo~i,loi, dQng tIT,vai),cuacaengu,..;haybi~udi~nQLF (Quasi-LogicalForm);...Cachtie'pc~n nayho~tdQngtot vai caeCallchuffnml,l'c,ng~n,nhungkhi g~pcaecall ba't ky/dai,thl h~khongduav€ duQccaed~ngbi~udi€n bentrongdo. 131 4.4.2 ;:, , ~ ~?~? ~ , , NGUON TRl THUC CHO Vl~C CHUYEN DOl CAY CU PHAP Xet v€ m?t ea'utrue,thl vi~eehuy§nd6i cay eu phapehinh1aehuy§n d6i ea'utruechill euacali, con xet tIeDb€ m?t, thl dayehinhIa bai roanehuy€n d6i tr~t1!tutrongcali.Khaini~mtr~tvtuphaidu<;)ehi§urQng1atr~tveuatu,eua ngu,euacaethanhph~ntrongCali[33].B6i vdi tie'ngVi~t,dothuQelo~ihlnh ddnl~p,Dentr~tt1!tu.la IDQttronghaiphudngti~nnguphapquailtrQng(ben qmhphuongti~nnguphaptuhu). Tuy tie'ngAnh va tie'ngVi~teunglo<;iihinhtr~ttVtu S-V-O (day1alo<;ii hint ph6bie'nthunhi sau lo<;iihinhS-O-V [31D, nhungnhinchung,tr~t t1!tu trongtie'ngAnhva tie'ngVi~tvankhaenhau[36],nha't1athanhphgndinhngy trongdanhngu[9].VI v~y,vi~exemxetcaenhfmt6anthu6ngde'ntr~t1!tu1a di€u ra'tdn thie'tvacaenhan_t6nayehinh1acaengu6ntrithuedungd§ ehuy§n d5icaycuphaptutie'ngAnhsangtie'ngVi~tdu<;)eehinh:xaehdn. 4.4.2.1 NHA.NTO VB LOAI HINH NGON NGU GreenbergnghieneuucaengonngutIeDthe'gidivanh~ntha'y:lo<;iihlnh ng6nnguvalO<;iihlnhtr~t tvtuanhhu6ngnhi€u de'ntr~tvvad?edi§meuacac thanhph~ntrongcau.Vi d1,l: - Tr~tt1!giuatinhtuvadanhtu(tie'ngAnh:A-N, contie'ngVi~t:N-A) - Vi tridanhtuehinh(headNoun):tie'ngAnh:sail,tie'ngVi~t:trude - Vi tricuagidituvadanhtu:tie'ngAnhvatie'ngVi~t:P- N - Vl tridanhtus6huuvav~tS6huu:John'sbook~ saeheuaJohn. - "I methimyesterday~'__~"Hornqua,toi(dil)g?Panha'y"(beta "dil") 132 4.4.2.2 NHANTOVE HINHTHAI Vi dl,lxettr~tt1!giua2 thanhphgnb6 ngutnjctiep(DO:directobject)va b6ngugiantiep(10:indirectobject)trongti€ng Vi~t,tatM'ynhusail: BIit fayvaotIii (+);BIifvaotIii fay(-); BIit banfaydaydaum{jvaotIii(+);BIit vaotIiihemfaydaydaum{j(+); C6nghlala: khailuc;1ng(chi~udai)cu~DO seanhhuangdentr~tlfcuano. 4.4.2.3 NHAN TO VE cD PHAp TiengVi~tthuongdunglai chudQng(activevoice),contiengAnhl~ithfch dunglai thl,ldQng(passivevoice),di~unaycangtha'yr5 trongvanphong KHKT. Vi dl,l:"whentheprogramis activated"-+ "khi ]dchho~tchuang trlnh". Tieng Anh chu yell dungdqngdanhboa (nominalization),con tiengVi~t th1 lqi thfchdung dqng dQngboa (verbalization).Di~unay cangthay r6 trong vanphongkhoahQc.Vi dl,l:"machinetranslation"-+ "djchmay". 4.4.2.4 NHANTOVE NoDNGHIA DanhtuchI mQtloai/thli.chungchungnaGdo,seduc;1cchuy~nd6i (tr~t tlf !chen/xoa)khacvdidanhtuchImQtloai/thuCl,lth{ Vi dl,l: LenngI;Ca/xuo'ng ngI;Ca(-7-) lenngI;Ca(; (-) ; Lenlung con ngI;Ca{;(+); nh(1cvanglen (+);vanglennh(1c(-); vanglentitngnh(1c(+);. 4.4.2.5 cAe NHANTOKHAc Ngoaifa, connhungnhanto' v~ngudl,lng(cautruc£)~-Thuyet),nhanto' logic,tam1:9,nhanto'vanhoa,muannha'nmqnhlxemtrQng,bi~nphaptutu, ... 133 4.4.3 , ,.',. ,.?,.? ,. , , ? CACH TIEP C~N CRO CHUYEN DOl CAY CU PHAP CUA BTL Sail khi xemxeI caeeaehtiepe~nkhaenhauelingnhucaeyell to anh huangdenslfehuy~nd6icayeuphap(haytr~tlftil), chungtoinh~ntha'ydog: - Neu chi dungeaclu~tco dtnh(nhutrongcachtiepc~ndlfatrenlu~t)doeon nguaitlf fighTra thise giaiquyetduQecachi~ntuQngloi, nhungkhonggiai quyetduQccaehi~ntU'Qngphl,l.Ngoaifa, conc6 nhii'ngslf chuy~na6ima chungtachuaphathi~nho?cchuaauatMnhlu~tduQc. - NeuheaDloandungcachtiepc~nhQc(b~ngthongkehaymayhQc)trenngG' li~uchuaauQcbeanchinh,thi<iQchinhxaechuach~ccaova thaigianhuan luy~nclingkhaHiu. " Chinhvi v~y,chungtoi <iffchQncachtiepc~nlai, nghTala: vila dungcac lu~tco dtnhdlfatIeDcac"n~anto v~lo?i hinhngonngG'"(4.4.2.1),vila dung mayhQcd~hQctrenngii'li~uciaduQcphilotichcuphapvalienkefcayehocae nhanto conI?i. Voi cachtiepc~nnay, chungtadamba0 vilachuy~nd6iehinh xaecaehi~ntuQngloi, vila giai quyetduQccaehi~ntuQngphl,lmataehuaco lu~t.Voi nguli~uhua'nluy~nngaycangbeanchinhbon,thicachi~ntuQngphl,l seduQcgiaiquyetngaycangtotbon,vachungcuQcla kefquachuy~nd6icay euphapsecaobon. 4.4.3.1 CHUYENDOrcAY cD PHApBANOalAI THUAT KFTBL D~GaplingduQeyelldu lai t?Ogiii'a2cachtiepc~nn6i tIeD,chungtoisii' dl,mg iai thu~thQeKFTBL - mQtgiaithu~tchophepke thiladgura cuamQt M/cachtiepc~nkhacvatiept\Icsii'a16itrenketquanay,d~choraketquatot hon(xinxeml?i d?cdi~mcuagiai thu~tTBL a phgn3.3.1.4).NghTala chung L ' toisedungcae lu~te6 djnhd€ gan nhanedso, r6i M sotie'p tue ciii thi~n tren do. 134 (j 113.'j,\)~i \.()il\ c,\\\l'j~l\l1&ic,3.'jc,\1"Q\\i"Qti~l\~~\\ ':>a.l\~c,3.'jc,\1"Q\\i"Qtl~l\~ Vi~tduQeduav~d;:mgbairoangallnhanV!triehocaethanhph~nellphap trong calltiengVi~tsovdivttrieuachungtrongcalitiengAnhgee. GBnhan ehuy~nd6i Ngfi'li~uhuanluy~nEV gallnhantli'lo<:ii,ellPhap' lngfi'nghiavachuyinii6i Ngii'li~ukh6nge6 nhanehuy~nd6i Caengu5n trithliev~ ehuy~pd6i tr~ttvtU T BQ gallnhan ngii'nghlacd sa """""""""""""""". .. .. . i Khunglu~ti . .. .' Nguli~uduQegall nhanhi~ntanh Caelu~t lingvieD Ngu li~ugall nhan theolu~t ling vieD Lu~ t t<5iu'u y '\. llfinh 4.18.M6 hinhhuiln'luy~ndE gallnhanehuyEnd6i cay enphaptrangBTL. 135 4.4.3.2 NGU LI~U HUAN LUY~N Vdi cachtie'pc~nmayhQclien ngfi'li~u,chungtaphaichuffnbi ngfi'li~u huanluy~n.Ngii'li~u($dayphaila ngfi'li~udadu<;jclienke'ttu, lien kef ngu phanrichcuphaprhoca2ng6nngfi'.Chungtamay~n kethliacackefqua- naytUcactftngtren,baag6m:lienkettu(4.1.1),gallnhanngfi'phap(4.2),th~m chieelgallnhanngfi'nghia(4.3)rhongfi'li~\lsongngfi'Anh-Vi~tEVe. 4.4.3.3 GA.NNHANcd Sd ChungWisad1:lngcaclu~t ehuy~nd6ico dinh,gifnli~nvdi m6ilu~t sinh dU9Csad\lngtrongbuckphanrichcuphap.D1,1'aliencaclu~tsinhdu<;jcsadl;1ng trongcali tie'ngAnh, chungt6i trabangd~rimlu~tanhx~sangcae lu~tsinhrho calitiengVi~t.Cl;1th~nhusail (dungki hi~ucuavanph~mITG ($phftn4.4.1.2): Bang 4.3:Th~dl;1v~caelu~tehuy~nd6i codinh. So Lu~ttie'ngAnh Lu~ttiengVi~t 1. s~ [NPVP] NP~ [ADJP NP] s~ [NPVP] 2. NP~ 3. 4. NP~ [NP1NP2] VP~ [haveVP] NP~ VP~ [eVP] 5. VP~ [alreadyVP] VP~ Vi du:cali tieng Anh: "I have already read that interestingbook" se du<;je chuy€nd6i (b~ngcac lu~tco dinh($Bang4.3)nhusail: [[I/T6i]NP [have/e ADJP book/euonsaeh>NP jVP>VP]VP.] ---- K€tqualasecocau tiengVi~t:"T6i dadQccuonsaehthuvi dor6i". 136 4.4.3.4 KHUNG LuA T Cac khunglu~ttrongbai roannaysesli'dl:mgcacnhanto'tr~tt1Jtutucac ngu6ntri th(tcco lien quailde'nvi~cchuySnd6ivi tri (ngoainhtingngu6ntri thti'cdadu'Qcsli'd1,lngtrongphfingallnhancosaa tren),vi d1,lnhu'sau.: Word Word LEN LEN LEN pas pas pas SJ'N SJ'N SJ'N NSEM [NSEM - NSEMa a1 a2 Word LEN pas SYN m , - N:~M]~ U No[Na; - Naj - X - Nak] ;,j ,k=O ;:;;:j:;;:k Tron£do: - Wordla bie'nthShi~nhlnhthai(nhu':"of", "account",...)cuatu; - LEN labie'nthShi~nchi6udaicuanode(solu'Qngtutrongnode); - P~Sla bie'nth~hi~ntuloi;li{NN,VB, DT, IN,..}cuatuhaynhancuphap cuanode{VP,NP,..}; - SYN labie'nth~hi~nvaicuphap {Sub,Obj,Modifier,Goal,...}cuanode; - SEMIabie'nth~hi~n hanngunghlacuanode - x: Iathanhph~ndu'Qcchenthemvao; - Chiso'du'oichivi tritu'ongd6icuathanhphfindotrongNodecha,voim la vi tri xanhattint ill d~u.NaIa nodethapnhatmabao(dominate)mQi s11chuySnd6itrongIu?t. Lu'Uy: Khongph,hmQithongtin Word,LEN, P~S,SYN vaSEM d6udn thie't clingmQtIuc.Moi nodeconNaiENa co th6chuacacnodeconNaixvoi cac thOngtindu'Qcdinhnghla(mQteachd~qui)tu'ongt1;1'nhu'Nai. Chungtaxemxet vid1,lsail(d6dongian,tabocacleYhi~uchisolIen chiv6tri thucdungtrong chuySnd6i): No[Nal[NalI -NaI2]-Na2[Na21-Na22]]=>NJNa22-NaJNa21]-NaJNaI2 -NaIl]] 137 Lu~tehuy~nd6iliencod,~lllgcaycuphapnhuHInh4.19dudiday: Na /1i1;( c; Na N." /'1\~i -~N I al T ~ NaIl Nal2 Na21, Na22 Na21 Nal2 Nail HInh4.19:KhaDangehuy~nd6i diy ellphapb~ngKFfBL. 4.4.3.5 - ~ ~?~? , DAY LUA TCHUYEN DOl DUQC RUT RA Sail quatrlnhhua'nluy~n,h~th6ngnayserllt ra duQcdaylu~tt6i u'uco d~nggi6ngnhu'nhunglu~tdudi day (chi tie'txin xemphftnthl1cnghi~ma chu'dng5): (POSA'=Qwh)n(POSN =Aux)n(POSN =SP)n(POSN. =VP) n 0). oJ 0.' Co nabla la: ""1 N N "d kh~" t>=>1\a2- aJ- a3- u'Qc ong ne'ula calihoicocayellphapngu6ng6m:trQdQngtu- chungii'- vtngii',thlno sedu'Qchuy~nd6isangtie'ngVi~tthanh:chungu- tr9dQngtu- vtnguva themhu'tu "kh6ng"vaocu6icali.Vi dl;1:"CanyouspeakEnglish":=>"Anhco th€noitie'ngAnhdU'Qckh6ng?". 4.4.3.6 , , ,~, ? ~?:! HAM DANH OlA KET QUA CHUYEN DOl H~th6ngse danhgiLlke'tquachuy~nd6ib~ngeachsosinh cayell phap dichvdicaycuphaptie'ngVi~ttrongnguli~uvangEVe. Vi~csosinhduQc tht,tchi~nd6ivdi caeky hi~uda'ungo~c(bracket)vacaenode.Tuy nhien,cae nodetie'ngAnhnay phaidu'Qc-chuy~ndtchsangcaenodetie'ngVi~ttu'dngung (thOngquam6ilien ke'ttutrongEVC) d~mdisosinh khdpvdicau tie'ngVi~t. M{Jtphftncaeke'tquanghienCUllv€ kh6ichuy~nd6icaycuphapdadu'Qcc6ng b6trang[72]. 138 4.5 BAI TOAN DANH GIA CRA T LU<)NGlIt DJCH EVT M1,1ctieuchinhmalu~nannaymuond<;ttduQcchinhla hl$dichtl,l'dQng Anh-Vil$tco chatluQngkhavachatluQng~d6ngaydmgduQcDangcao.Nhung lamsaodaubgia duQCchatluQngdichcua n6 ? TITtru'ocdennay,nguaita thuangdaubgiab~ngray.Nhunglamnhuv~yseva rungtonkern,ch~mvac6 th€ khongkhach quail (neuchinhnguai xay dl,l'nghl$thongtl,l'danhgia). Ngoai fa, neu danh gia b~ngray nhu tren, ta se khong th€ danhgia nhii'ngthay d6i hhg ngayva thuangd~ndenhil$ntuQngflip-flop(dayIa hil$ntUQngmakhihl$ th6ngc6mQtsl,l'thayd6inaod6d€ kh~cph1,1cmQt16isainay,nhunghl$thong sed~nden16isaikhacmatakhongngatoi).Trai l<;ti,neuchungtac6duQcmQt GongC1,1danhgia tl,l'dQngvatucthai,chungsegiuptahi~uchinhhl$thongngay l~ptucvad~dangchQndU<;5ccacthamsotoiu'Urhoh~thongd~ch.Ngoaifa,voi ketquadaubgiatucthl (on-line)nhuv~y,chungsela ngu5n.dQngvieDratJOn ehocacnhaxaydl,l'ngh~dichmay. B€ giaiquyetbairoandanhgianay,chungtahaynhlnl<;ticacbaitoaDda giaiquyeta nhii'ngphgntruoc(nhu:gallnhanhlnhthai,ngii'phap,ngii'nghTa, ehuy€nd6i),chungtad~udunKhamdaubgia Scorenhutronggiai thu~tTBL (bhg cachso sanhnhandu<;5Cgall voi nhandungtrongngii'li~uvang).Tuongtl,l' nhutren,chungta fighTngaydenvi~cdanhgia b~ngcachrho h~chungta dich caecalltiengAnh trongEVC, r5i saild6sosanhcautiengVi~tdichra cuahl$ vdicalltiengVi~tchu~ndaduQcnguaidichtrongEVe. Th~tv~y,vaiDamg~n day,cacnhadichmaydalamnhuv~y.NhungquailtrQngla sosanhnhuthe naG,khongth€ sosanh1-1duQc,VIrungmQtcalltiengAnh,chungtac6th€ c6. nhi~ucautiengVil$tkhacnhau(v~cu phap,v~tr~tW tu,v~cachdungtu,...) nhungv~nduQcxemla dung. 139 Xuatphat tuytu'ongcuaKeh-Yih Suva cacd6nglacgiatrongc6ngtrinh [139]v~daubgia h!$dichAnh-Hoa,chungt6i duabai roanso sanhgiuacali tiengVi!$tdomaydichvadiu donguaidiehv~-bai roantlm1<)trlnhngannhat giua 2 di§m: tu di§m xuat phat - R va di§m dich Q. Trang d6 R = {CII'CI2,..., Clm} la cali tiengVi!$tdo maydichc6 m tu (tITchinht3.)va Q ={C]I ,CI2 , , Cln} la Calidonguaidiehc6n tu (tuchinhra). C2n End t6i C22 C2l cua Start Cll Cl2 Clm Blnh 4.20:L<)trlnhdi chuy§ntu caliR dencali Q. Chi phI duQctlnhb~ngD =Wd *nd+Wi* ni +wr * n,.+Ws* nstrongcI6: nct,OJ,nr va ns1aso badexoa (Del), chen (Ins), thaythe (Rep) va ciao(Swap) tu'dngling (NOP:kh6ngthayd6i) va Wct,Wj,Wrva Wsla cach!$so tudngling (h~ sonaytuythu<)cvaong6nnguva chQntheothlfCnghi~m).Vdi wd=l,wj=5, wr=5va ws=6. Tu R -7 Q, tac6th§e6nhi~u1<)trinh,vacacchiphisekhacnhau,vi d\l: - / / nayLL' / I I \t .'. tinhl I..".+.'. mayL R day la may tinh ? t6icua R day la___- may tinh ? t6i tac6: nd=l,nr=3cua Q may tlnh nay ? t6i D=lxl+5x3=16cua Opl REP REP REP DEL NOP NOP (duangnetdlit) 140 Bai 1<)trmhtren(conc6th€ c6nhi~u1<)trlnhkhac)tuonglingvoi2d6thikhac nhautrong Hlnh4.20henphai, trongd6:duongngangtuonglingvoiDEL, duongdQc: [NS,duongcheo:REP ho?cSWAP. D<)chinhxaccilaCalidichdu<,:jctint bdi: l( =m~. voiDIDiola du&ngdi ngiinnha-t(chiphitha-pnha-t)tuR toiQ vamla . /Lmm lhi~udai cali.Vi~cxacdinhDmioduabai toaDv~bai toaD"Nguoibanhang" ), lienthu<)ctrongky thu?tquyho?chd<)ng(dynamicprogrammingtechnique). Tuynhien,theecachgiaiquy€t tIeD,tathayKeh-YihSu chIquailHimv~ ~tkyhi~u(tuchinhtit),chlichuaquailtamd€n nhungy€u to ngonngu,nhu: U,ngu;nghlacilatu;...D€ kh~cph1,lckhuy€tdi€m tIeD,KishoretrongGongtrmh LEU(BiLingualEvaluationUnderstudy)[96]daduaray tudng:thayVIchIso sanhlienlilli-gram(l tuchinhtit)nhu Key-Yih Su,thltasf:sosanhtIeDn-gram (ntUchinh!ii, voi n=4,~hl"maytint cila toi" trongvi d1,ltIeDsf:du<,:jcxem nhuJ L I Count clip (n - gram) Ce(CulldJale.,' }n grameC" adnvi)nhusail:P n = L I Count (n - gram) C e I Ctmdidale." ) II - gram e C (4.19) Voi Candidatesla t?P caccall dichb~ngmay,conCountclip(n-gram)la nhungngu(g6mn tuchinhtii)xua-thi~ntrongCallngu6nvacalldiehtrongngu li~uvang.Caiti€n naydatinh_~€ndonvicaohontuchanttit:d6la tuvangu, nhungchuatintd€n ngunghIacilatu(cungm<)tynghla,nhltngc6th€ dungcac titkhacnhaud€ th€ hi~n).VI V?y,chungtoi khongsosanhtrungkhop,maso sanhcoclingho~cg~nlOpngii'nghlahaykhongnhutrongph~n4.1.1.3. R day 1a may I tinh ? toi tac6:nd=2,nj=lcua Q may1 tinh nay ? toi D=lx2+5x1=7cua I Op2 DEL DEL NOPjNOP INS NOP NOP (duongnetlin)

Các file đính kèm theo tài liệu này:

  • pdf5.pdf
  • pdf0.pdf
  • pdf1.pdf
  • pdf10.pdf
  • pdf11.pdf
  • pdf2.pdf
  • pdf3.pdf
  • pdf4.pdf
  • pdf6.pdf
  • pdf7.pdf
  • pdf8.pdf
  • pdf9.pdf
Tài liệu liên quan