Base_JSON/prefixC/json/C88/C88-1014.json

{
    "paper_id": "C88-1014",
    "header": {
        "generated_with": "S2ORC 1.0.0",
        "date_generated": "2023-01-19T12:15:08.617138Z"
    },
    "title": "PROBLEMS OF DESCRI~I'ION AND RESTORATION",
    "authors": [
        {
            "first": "Igor",
            "middle": [
                "A \u00ae"
            ],
            "last": "Bols}l&kov",
            "suffix": "",
            "affiliation": {
                "laboratory": "",
                "institution": "VINITI~ Academ~ of Sciences of USSR Moscow",
                "location": {
                    "addrLine": "\"i252\"19",
                    "country": "USSR"
                }
            },
            "email": ""
        }
    ],
    "year": "",
    "venue": null,
    "identifiers": {},
    "abstract": "",
    "pdf_parse": {
        "paper_id": "C88-1014",
        "_pdf_hash": "",
        "abstract": [],
        "body_text": [
            {
                "text": "ABST]~CTo Russian elliptic const~-uctions are exami.ned from the point of view of syn~ tactic analysiso Reciprocal elements in a co-ordi~Ltive elliptic sentence are exposed and possJ.ble types Of the:i/' similarJ.ty are explored. Linear formulae of el].ipsis for most textual cases az.e constl~cted ~md statistics oJ~ their use is discussed, As a re~ sult the main steps of ellipsis restoration algorit[bm are outlined, :[NTRODUCTIONo The investigations of e]_~ ].ipsis ({;apping) in natural language senten~ cos with st.~/ctural methods have been car.. tied on Tot more than 20 years, but a]go-. ribhms oil automatic restoration of omitted words either in coherent Russian texts or in sequence~ of Man~-Machine interac~ion replicas have not yet been proposed\u00b0 The prob~ lem is stiil tepicalo Indeed, at an average each 7th entry in Great Soviet Encyclopedia and 25th abstract J.n a co,muCh Soviet abstract jo[,m.nal contain al; least one sentence e~ a folJo~i.ug kind: ~0 BOOM oTpaonsM 9HdIIM~ ~0 BTOpO~ -rL'0~EO ~0 x'y~aH~TapH~M Hay~aMo /The first database contains information on all knowledge field s~ the second one ~ only on the Eumanities./ Conversion of such a sentence to a for-, malized language, e,g.for automatic updating of lactograpkic databases~ without the ellipsi.,~ restoration is impossible. ~eenwhile ellipsis in R~ssiaa~ sci-tech -texts is very diversified and covers sxqV pa,'t of a sentence and most frequently the predicate with ad~iacent words\u00b0 Earl~ Soviet works /Leontieva, ~965/, /Paduche'~a et al~ ,q97~/, /Korelskaya etal., ~97~/ had examined the phenomenon from the point of '~iew of synthesis. But when synthetic rules transform a non-empty entity to an empty one~ corresponding analytical rules are not their trivial reversion. We examine elliptic constructions in co~ordinative sen~, I;ences w:Lth orientation to analysis ~ ioeo to parsing algorit~u~s including restoration of omitted words\u00b0 Several importa~t issues should be forced for our purpose~ 1) introducing the no= tion of :eeciprocal elements in non-omitted parts of elliptic sentences; 2) defining new types of s~agmata for restoration of semi~ destroyed links between words within the re~ ciprocals; ~) exploring possfble types of reciprocal similarity; ZlO describin~ co~-or~-. diz~tive sentences with minimal n~$er of elliptic transformation formulae; 5) collec-~ ring statistics of the formulae use~ which implies a search order of a specific formula for a given sentence\u00b0 Thus a base for outli-~d.r~ the aain steps of an ellipsis restoration algorithm is formed~ RECIPROCAL ELENENTSo There are several kinds of word omission in natu~'al languages. Among them co~ordinative reduction is car~ tied out according to the formula XQ & X~Q Co~-ordinative ellipsis includes co-~ordina bive reduction as a subset. Two or more pln-ase segments co-ordinated in a single sentence have some identical parts not necessarily standing at [;he borders of the segments. ])'or economy's sake the natural ].ang~zage omits, wholly or partially, the re~ current part of a secment , ~s a rule, in the second or subsequelrb one. The omission may be carried, out according to the formula XQoY & X~Q Y* => XQ~Y ~ X ~ -Y*, where X and X\" are u. p,.~ir of somewhat similar, but eom~ pare~&nd even oemantically opposed elemento referred to hereinafter as reciprocal; Y and. Y~ are :~nother such pair; the co--ordi~Latire segments are XQ Z and X* -Y*. In ~Lussian writing ellipsi~ is often accOmpanied with a dash.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "The colmnon feature of ellip~;io omissions is that a connected dependence tree for a sentence cannot be built without their re-storation. The parsing algoritbJn should i]akc into account and [~roperly distinguish the specific phenomena occurring in a given sentence~ e,g. ellipsis and zero copula going together\u00b0 BASIC ASSUIAI~TIONS. We examined a co-ordinative ellipsis asslm~in~ the \u00b1'ollowinc:",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "A context confined to a standalome el~ liptic sentence is sufficient for resto--r~tion.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "An ellipsis in a sentence is explica~ ted. with a dash (along wi~h some implied in~ dications) ~ but omission may occur not where the dash is or not only-[;here. Comparative and other eo~stl~ctions without a dash have not been examined, but most of our sea-bements cover them, too.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "Omitted parts, after restoration :~id, if necessary, morphological re-agreement, exactly restore the original mea3zing of a given sentence as a whole\u00b0 Therefore recurrent parts in different segments have had exactly the soJae length, linear order, and lexical content before omission.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "i dependency tree was selected as a tool for description of links between word-for~ns in a sentence, and the nature of the links corresponds to those suggested in \"Meaning -Text\" model /Mel'~ul, 1973/.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "Co-ordinatively reduced, words should be restored, along with the ellipsis proper, where @his provides connectedness of a parse tree and saturation of obligatory valencies.",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "There exists an algorithm for syntactic analysis of Russian texts which can cope with parsing any ellipsis-free sentence. It can also parse, consistently and unambigq/ously t the complete segments of an elliptic sentence and tie links, even a~bigucus, within word groul~ in incomplete segments preparing them as dis jolted \"bushes\" for a final parse tree\u00b0 STATISTIC OBSERVATIONS\u00b0 As many as 600 elliptic sentences were extracted from large (more than 1000 pages) corpus of texts in various fields and genres (abs~ tracts~ articles~ popular science books~ and brochures)o The material was arranged in a miz~imal number of elliptic formulae, A total of 2~ formulae was found, but only four of them exca~ed the ~+% threshold (see Table) .",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 727,
                        "end": 733,
                        "text": "Table)",
                        "ref_id": null
                    }
                ],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "In the Table the arrows show the direc~ tion of synbhesis~ Qo stands for an omitted part replaced with a dash; Q ~, Q~ are ad~ ditiorml omissions posoible ~6 th~ left a~.d to the right of the dash; P and P~ s~e op: tional segments (modifiers ~nd the'llike), not involved in the ellipsis\u00b0 Available statistics permits to cone= lude.",
                "cite_spans": [],
                "ref_spans": [
                    {
                        "start": 7,
                        "end": 16,
                        "text": "Table the",
                        "ref_id": null
                    }
                ],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            },
            {
                "text": "-The most widely used formula (single omission between two reciprocals) has ex= ceeded the use total of all other formulae~ and four leading formulae together account for more than 80 per cent~ Number N~of reciprocal pairs and number N of o~issions in a sentence us,ally satisfy ~he formula I N~ ~ N~ I <= I, since reciprocals and omlssiofis ar~ commonly in~ terleaving each other, but it is possible to construct courtier-examples. Approximately four per cent of all cases did not fit into our formulae. Either a possible formula was too complex to be practical, or the shape of the sentence was dubious and even incorrect from the point of view of a human editor, ENCLOSING SYNTAGNATAo Before searching reciprocals it is necessary to establish syntactic links within word groups in an incomplete segment. The convenient tools for deseriblng these, links, the so-called syntagmata, are not 'always sufficient here~ see, e.g. the word combinations: odQd~eH ~ HNO 9EOHOM~qOCE~O/0T~O~BHblO Hp0HsB0~C TBOH~:. ",
                "cite_spans": [],
                "ref_spans": [],
                "eq_spans": [],
                "section": "",
                "sec_num": null
            }
        ],
        "back_matter": [],
        "bib_entries": {},
        "ref_entries": {
            "FIGREF0": {
                "uris": null,
                "text": "~> X & X*Q or QX & QX* => Qx & x*, where,X ~md X ~ ~-~e different words or word groups Q is a recurrent group of words, and & stands for a co-ordinative conjunction or just a comma: (He took)Q (the bread) X (and)& (he took)~ (the Imilk)x. -~> (He took)Q (the bread)x (and)& (the milk)x,.",
                "type_str": "figure",
                "num": null
            },
            "FIGREF1": {
                "uris": null,
                "text": "Cases with N~> 2 and/or NI > 2 amount for less than two per cent.. Approximately in 7% of cases nonsymmetric (\"chiastie\") formulae occurred\u00b0 For example, formula P,~Q, XT & X*QY* :> P,~Q,k'T & X* & Y*may have t~e following realZzation~ (C0~MeCeHa~ pado~a)p (~podyo~)Q@\u00a2 ~mi~m{opa )X(pacm~peH~ Ma~oMaT~ec~o~o Epy~osopa )X (,a)& (o~ Ma~o~a~a )x* -(~na~o~ H~ HOEoT0p~IMM MaTOMaT~qe0E~E BHaH~M~y~ o /(A joint work)p (requires)Q (from engi~ neer)x (broadening of his mathematical outlook)y (and)& (from mathematician)x\u00a2 (the possession of some technical background)y, ./",
                "type_str": "figure",
                "num": null
            },
            "FIGREF2": {
                "uris": null,
                "text": "n~o(HoEa%a~ea~) . /.generalized economic/ some proauc~ional kinaications)/~ 17 eta o~ p~I/H~:~m HOBBIMM (npo~p~l~TZ~I~)/by 17 old/by five new (enterprises)/; E8 EanMTa-n~c ~qecE~x/~s co~aa~c ~qec~x (c ~paH) ./from capitalist/from socialist (countries)~ ~cao m~e~xcH/u~cj~o ~EO~ H~0dpeTeHH~X (EH~2) /the numoer o1 available/ the nhmber of newly bought (books)/; 80 TNC~ n~-dn~c~ec~x/20d ~Co Hay,iHmx(nyda~ KaEM~)/80,000 publicistic/200~000 scien~ tific (publications/. Reciprocals in the pairs above are divided with slashos~ ~A connected nov_us (derived fz~om complst~ seg~merits) 6u~e given in pa~:enthese~oWe propose dependencies of a new toype~ named enclosing s~ntagmata (E~) ~, These ~,z'e attributes ~ li~in~; br.othe:~: o nodes :L~:ou ~ight to ].eft (O~O6~OHHNO ~ D~()IIO~HqOS~:~iZ:~)   or between a numbez 0 ~_a~d oo~,suboz0di -, nated attribute (17 ~ (]TapNx) o:~: beb~ee~ a proposition 02 a no~ wi~h pwedlcati~ p:cO~ ~ erties and subordinate ',o_Ot~zl ~ ~ att~rlb~te 88 ~ EaEm~aaMc T~qOCI\u00a3MN:~ 'qMoJio ~' \":-~ !~MO:SI~!/fZ:C H Several ~S constitute the :Li~k~~ 'bo~bv~ee~)~ a dominating word aud its indi~:oo'Sl~' sdbo:~bdinated one~ Via ES the no(\u00a3es ~os'b i~<@O:<~t~:~ ant fo:c reciproc~G matching b~!~D.ed tO be within easy d:iost~ce fz~om thei:,:, b1:~shos ~ l'OOtS (may be~ at She V{~y ~OObS),, Simi:Lo.x ~ links can be established in eom~!ote seg-~ ments~ too~ They transfoz.m a :col<)v&ut ~:~b ..... tree to ~u~ acyclic gzoaph~ which :~'aci:l.iba'berJ matching nodes a~d arcs\u00b0The set of ES is sdbject to ~:~:f,dating so fa~.o Caution should be used. howeve:~ dbo'~'b updating~ Indeed~ atterapts to direetly ].::i.,nk words arbitrarily 6Jostant within, a eonvo~:li .... ent~depende~ew t~-ee~ bheug:h eliminating /the vex~y notion of ellipsis~ lead to s~!perfl~ ous comple~ty of ES an& of a global rood.el of natural l~guage~ too~ SI~'~IARITY FF&TURE, S~ ~.~,~u~l seg~enta~-, tJ.o~ of all available sentences has made clear that antagonists in reciprocal pairs are not in general ease m~tually isomo~ ~ phic~ i~e~ their subtree do not qDite c~m .... cide~ Therefore, the labels at the matched nodes shottld be forcedly involved ~nld these are of the following t~pes~ Lexicalo Lexemes at the roots e~d/ol o thsir~gt subo~.~linates in 22 pe,\"~ ce~fi~ were strictly ooincide~ ~orphological~ In most eases le[~emes at the ma~~ 'belonged to the same pa:cbof speech~ and their wordforming chef'acted: o, istics were in agreement~ nouns and nu~be:cs expressed by words ~ in case~ ad~ieetives (pronominal included) and participles ~, in gender~ number~ and case; personal verbo~, forms -in gender~ number~ and pex.seno 0nly in 30 per cent of cases the agreement h.a~ not been revealed (numbers in digits~ ab.~., breviations, etc o)~ S~q.<Some indlcato~s bz~eated. :b~ the ~~~ Text\" model as synbactie~, ;~ight coincide~ e~go interogati~J;oy of lc,~ xemes ~Y~i and CI{O:~BKO in the ~.~tenc~ H~o /I as~ea where ~ should go~ and D.6 asked what time it was\u00b0/ Semantic\u00b0 If matcld.ng labels of bAo three-~mentioned failed oz ~ at onc~ se.-~ veral nodes in a complete segment ~e~:~. ~ s:h~,~: ilar to the node in ~ ineomp/.ete ons~ [;h~,~. coincidence of even one sema~)J;ic :indioato:c is importsnt~ Taxono~%, of such indio&to:c~ is not establisked ~eto Several J','o, oet.~J (classification aspects) wlth a~b~isslble intersection of thei:e scopes a:m:~ e:~p~oted to suit well~ b'~,t slm:ple thes~/:s,~ioal hi,~3~eitho:co 8J:ohies a~8 not excl~,ded~ ~<9 hav<~ speoifZoally observed~ .%%~antZi;a~:bIv<~ a%~,~, oar&inai words (28%)~ synom~V asd, ~'x~O~Ti@' ' When semm~tic simila.~:it~y w:'lt~CL:a Oai~ X/X* and Y/Y\u00a2 failed~ se~m~tlc propo~.~t:Lo.~!~ X/Y ~ X~/Y ~ has been ~ome-bi~,.e~,:~ obrJ<~:.v<,6~,~ 'b<(!/i; ~ '(, ~O'~:d '{;}'~ \" < '%.: :~.;i+'i~j.<~.j.-{7 IJ t~: i ~ Oj~ ll;[i*(~ [JT:i;<~.~lll}~ ]:~; \" \"1 {;:i.o:~',~ J..lJ ( '~.'{/J:>J,6' <.\" ;,~ ]J!{<l?\",t) A,Lt;O}2'.i.tl!l!~:7~> .A.:~ a;kgcxe3.bZi.,.m wm ]:,:~.:t~,., )90~.~o :Co:\u00a2 :r.~-Jsbo~;o~tL<!.o'.r~. OJ~ t,,o, ,Jw0.'J~'~.tJ.b;/v'e o[!.,~ [I.7[pL~.o (0]~1) oo)'~f;:!~:~'L;~', o'< \"b]:,.o !:'<.)].'!.t;azl'}.l]{5 )0u.{].)7. ,u'i:;o ]~i[i g q ) J},, ])'u':ViO (\";o].).:;il)t.;:\u00a3~7' ,:i\"eom ~ ) a{].{,~tjs.<Lls]:l)!t j,f] C,>ll'hr?,y ;],,)_,t {i i O~ i. ~i}~.OJi&t(f~;g(_~ [J~Z~lD~,>g9 fJO:,.~'~<0-{3\u00a2 b~,' ,<!je:~ri;o>x:~co 0;P',~.:,f].ag :'Ls s~\u00a2o.].ba:o.eca:u::d.3< do. -~,~i'9\"@t]. 3!'370;~4 \"{;]\u00a2.] i;l'v~.<~t-) <t)'i~t\".)i\" '.'.L~!7.)>~ 87 {gt,~.[_.].:} :%\" 7 (}{}).~t,@0\"150(~'1t.' }[Jl:l OJ'.' '~]~i.O ]Jl.t)?/3~.} \"iJ}/O() g {JO\" 'O:C'(iJ.,:~-O.~' ~.Ti]_WJ.~.;>' OJf' 'gJi'qO ~:3@.'l],'[;Og].G\u00a3t 8\"0;ei*<$1;LXX,O 00;~]\u00a3[ ~)>(a,-: f{t(J~[~J{J {)f [< [~OM{} a {30[{)'{{\"!.{1.o ;{) 3.::i 'b:!:~o 0]t',.<CJ~!G .'7.s ~:Ioi:~ :i:o:,:, b}:,4:, ~~o;',,. 'TeZ].O@ '[;]/!,(J){{ O.[J,{\"O72:~_/b]\u00a3.~] ',J~' c;il,].J]i),.<JJ,f] ).%.}l~'iSO:~;tb, 'lTJ..o:O. tV[;Lt<Zo<{;H ~:<[f;o {'~'J,ts'(7~HbO:7dc<,IH ~<i ' > \"{;]7; :D)').OO).+, ]TZ{(9b%} ~Jc-]{gli!l;~).]7 i t]~l~i. {,{ li.ti.}l~tL ~)O~:{.'{;J.C)f!~j ;!(J:' ~_Yi].(Tt y<'O t3],G'ZdO:,Y[;hl :,~,+0 c<>t{'\u00a3UT!,O'{;tD~!i~, t6bc;{y ;~]~O'tfl.(]. 'bO :{.~waz;dL?f ).:L:,F[,.:,_;d. v!x~. oo,,v'v:,r\u00a3u>.ij oo./;.,s J.< <,.<:-. G@GI3>-}.]:,y' 9 (i~lC].OS:!,?d~ L;y/f~6,4.~)~ttY[;~o },~) {t}.]f!.~J ){]'J ~I1\"}, y+:' [g(]g'[]{[b, \"g~I}.{:;:} :: :t'(](~,:!,],~[,]~{~ 6EL,'4 .' ;-_< a~d. 7 'L:~!. \"b}'to ~O:,:ql]xJ'{so floGme;~:cU ,<-v,.o(:; f3(Tr]),~o,l'3,ot:J, tiiTt.t~'i/J.z!.[>~ .~:f?O)7! \"],iJi~ 'bO;~?(].@~?ft j.iiZ~OZtY\u00a31PA,, 9!]\"!.o y).od,l>>'bT<.o..to([,t~ :!ii~9,~zj]K~2nf~j J]o37 D]l=O })lJFJ~%e~3 f}'~ t] ~,\"0 FJ L~';~f' (}~ {I, \"[;.t [( }[i, ? ? [0() ~.)*171 i 9 \"t/If;J () j-t {J \"t. [I ] l(J(~,~} ~'~}\u00a3).\u00a2 ; ]JO].'Ti s, ~_tY!t't. i;]~.Ot~].({ be ~x~d;J.:o:ood. s.'i; ;Td,~7o:u 'b:,z~;o }L(;'\u00a2e], e, J,J7 ;,9~,g~.oi;:io:~. wo:/v\u00a2.],<.~ ~.,:\u00a2o OL, tt~otn;rbe:~_.~o(! El.; \u00b013I}<o .~.'OO'[Wi <. &) tlfh.<~ X .<t:o.(i \")f :Uo'~'nd.~, d.o]:)m:o.d.o:~_oLy ii_,f{!'~ .q_.~~ bb.e ool~r!)].oi;,,~ r~eC;me,):~b t,,.:~.ie7 l; ~ j.(Ji [~ o {[ 9 ~tttti,.~!.[[.[y\" i:O. ES(}(;:_r)ft\"i,]!{~ OZ@~D?v {i!]:te:; Si;i:~*iflJtX<' ~jY!.(i. ]..'JO(\u00a3t;J,o!t OJ',' \"}}O<\"t,'(J,(9)2I:~ \"bt)'[]W'O{~,t :L'O(~j,]])Z4~t:)(;~-j-!,fl {l=t't(j faJS!J.lJ[;J>-) (KO.~3 ~'.~.'~-{,) \"[;]i.'l[f_j ;x:O'v'm~&],@0_o ~,';) {I}}l~e ,q:{~ooJ.ZJ,cs ([t%:<i]o:r:,,ro.].~\\ :14J (Vi.~:t{gi}.O~;jo{{, O.L,<I.U, Oll!J.~IuO~L ~70)f\u00a2]. [,;%+OT!.]7~1 Q]/O @tlDJJ2],OO.0e2(J~ wibh J.;b8 ]:-to),Yk :U)2om b}.~.o ~$os~(tf!x;'be segme,-H:; i;o b}s,t; ;bJtoom])[t.obe \u00a2:m.Oo ]2_o...tt{4~:rteome~{; oJ:' Olin :,.\u00b0<~sb~ rue.f!, wo'.vSs J.s o~z:\u00a2/x':Lod, o'c!J; a \".i.:t ~llgC,@Sf382L-),, b Ool~o J.)'l :rl~O~gb0:~'~ 0 [pt-J:i~t~JO:O.~) ~:O_(t./Oa? I~e)-&&@Z' JTozu' & )))?O ,-~ di.6abo o ]]J:' /;b.G:Oo, o.V'o m.o:~:O L;])dzT). {;wo oo.~o;\u00a2d:i.L~z,,~ J;4.'C,'~ 130ft;l~tO))/[;f] JZfL 'IU]~O LIO]r.d.io:tl.oO 9 j;.]xO):/ \u00a3gi;@]]fJ } [JO !j ,'::it.'() :,20])6&JD\"@,{[ :[?OZ= (]'9e&#,~\" S'ctbrJeq,tte:rti; SOp.;m(;:,:'ti7 <>p'pl;y.Lxleq l):\u00a2eoedltxl G ( oom]Q.].ei;e o/p t~b3t, .r<,'3 ,O .td.) O[Q(;o Ag, Ttg0~t4!';lil])(:~']}llVqg.q-!L~Bo tt)ks.O ~i'O.'{;ib.o~,: WJ.~t~;9;5 -{;O o:~f;[i):ve~?i~:i ~lJ[.<~ G;L\u00b0e a-i; t:Q2EbJ.I;1X~e [;O l[ o (J o]}J.dLtD_,' ])oZ~oXo:m({:L~:~ ):,<~(?,..,dTl%bJ,ss],i.~_~ Ll~ogo24d.~,El:,.evt% ,'7:').0. OS}l,~oia.].]{7 i;O ~\"tL,]JoA~):eesJ.a.n. :[-'o:~! /;].s.oJ.:!: \u00b0 ]\u00a3o(:?.~ 6\"n:!.i;J.gp.e~ ]l.o ;].s ~ffl.,SO -i:;]c_tat!A-i\"r~_]. ~;o }VI,V~ )}':~7u.d(<:7:>!. onil. _A.,.A.~,].{<u:f]]{j,:o~o. :Coz ~ od.:i.%o;o:i.a]. :uoo, )7]~ J;w]Y.e J o ) @'dP3 i][0!B!f0 ] t',} i :!:::!u:~d,:! o'v4 ,j}fo]t'o (\"i e)(i!)) ~3~!.a'Ly,qis .:n~d. s2/sdi]L!.t'.;d.rl C, j7 ~tl]t.~!L';j,~:)'~), \u00a2.?],[l,J,]l)'0J,(J l:}(~:([b(-*Ll.(3O;Jo ~' JT',~l.'O.G]])l~l) .... i;o]e%:,]:i.ch.e,gl~:cry& J.:~rifo:fzm.&i;~jj.;yo.> \"I \"1 9 )!/{O~l(3OV0 ;\u00a2 vsR:<eT< ~ Z0 o 4.1, ea.~;<, )/; td./V;~}.%\u00a9 V;~..~ ]#L, ~f <, &+LG4, a fs]to+]l.e;t~tlo ~ {~ o }{L ('t c)'/15) \"r ]ii] ,:-, ].;7D'IJS aS '~,e;,_?O a:~!~])]_~O,',L\u00b0J.(~ ,~3J,[.!;:i),~ 0 ]'.'{[E!;CtoIhl'7,O,, 7'[osoow~ ?]:N]12:1: v Y@o 20 '~51 :Ko:ee].<~ka;Y 9. ,a :!! ~, ]:) o &;!?a.~h] clo.e'v'u. :t,] o Vo ( \"19 75 ) : t!!:e: <,~-fJt\u00a3O:i!ii!{S,'{;:LO]t ;til f-J,ygllll,()'t;.Z'ZC,. CO~2{h]Yi~,Ob:!.O]].L\u00a2~ C.,O--O:L(iill.OJ;J.ol~ t{*l({, e].f[~[[J,<JJl.so q ]~lo&-!.Gh~io-,i;e]dbilj... o]_~.e, st<a:y'a_ i)~:ffo:c.n~u:tbsJ.ya~go:~%299,~ lt~oscov~s \"V.I:N~[U:L:; P]2\" 29\"'38o !'.~e]. ~ ~,nk,~:E oA.o l,]:~i~o:eienoo i:c~. l;he b]~.eo:,:,y of }.to_,<.~siar0o )!;(osoov,a J~ro:aka~ 197}$o 2.A]-),!rat<, :!!.htJ :i'o'o',.' .mo~:/i; :i2~:ogt'<J:,rb O1].iLpsis :i:o:\u00a2m,:l]lae ............. :~-% ................................................................................................... : ............................ 77 ....... { ..... 7--, ...... ; ..... ] .......... 8\u00b00 [~:,o]~:0,o~q,~ ~,s :~\u00a2\"q,o~\":'t)q ~:> [b:,!o:~xq,<~yo i,:<,, . .. ~ -<,:,..., v,-,_. [~!,o[~.,%o:~< 8~ :e t .... :~!:>' <[)Vl (',:K]BIa:?m o0I)a ~-,a-f,lx)v..., \"' (~,/:$).< ( m~@ogBl~'l ~omJx~vJTo m. lae)ql ( ,:,a)~<; ('i:M}>c,\" (]; ~:..~<. ;$,,~]m-:,mm~m~oo teofi )-\u00a3,, ( ]'IOZ]TI b)0b1IIe])a \u00b6['yIJ~'~ 0;d~iIS, II;0]}8 (:~ -_1:98\u00a3 :,~o ~,:~ (@oW?:Q Qq 9!b{)..,, (,, 8+)gb ( 0 l.(0!l,~4aJ{:l)ll!,~:1{):7:m ~. (~VJ:b]i.o't:d; z'oundin{~; O:if}).?o(V/hc:a. ;JbJ.i\"[,:[~l[~ ad.dJ.b'J.o:rmt].):~< (<~o(le;J m~. o:uz'o:e will De) 0 (nega/,:L\u00a2o)-x ( :, m:).4)~ (w:b.(m ~dJ..:t:b:L>-G :r)Ovo:ese)7:o \"' (:_porJiix].ve)_~-~ (2/.3),~] (OJ~ mio:coaom._'pu.-be:c's ~ 'c !.~:i.<I.e}'c~.. l (o,~4)<; (1/3)~:+ ,, .@D:~'i.(\",&]. \"lT]:!.e:fl~).,~(i)].g:tFjNa j.,,-~ \u2022 J~ O-n~r.o4) Qo (iso~ e:c~8.o):\u00a3 (a<uT.)& (J.Z 07~])OSJ.'I;O oase))?q<-, (no:oJ.s o--Ise~,m:Lo):~o t\" c#o:t.?ea:%;!/uT). Z {.!.:Lo:<'u:~:o.e~ ~(m~ru:,~-O J:o:r,),, ' (600,~ 0()0 vo]. o )v n.a)~S (\u00b0 l: <.Oe.e,:i.a !.\u00a3 ~c:d.):\u00a3. \u2022, \u2022 (tZ~?lT,:eo:~dm,.~ge].~g ~I0 ~ 000 'v'o:k <, )X ':'",
                "type_str": "figure",
                "num": null
            }
        }
    }
}