O0~ORDINATIVE ELLIPSIS IN RUSSIAN TEXTS -~ 
PROBLEMS OF DESCRI~I'ION AND RESTORATION 
Igor A ® BOLS}L&KOV 
VINITI~ Academ~ of Sciences of USSR 
Moscow, "i252"19, USSR 
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 stat- 
istics 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 rep- 
licas have not yet been proposed° 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 abs- 
tract 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 upda- 
ting of lactograpkic databases~ without the 
ellipsi.,~ restoration is impossible. ~een- 
while 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° 
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 synthe- 
tic 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° 
Several importa~t issues should be for- 
ced 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° Thus a base for outli- 
~d.r~ the aain steps of an ellipsis restora- 
tion 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 
~> 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,. 
Co~-ordinative ellipsis includes co-~or- 
dina bive reduction as a subset. Two or more 
pln-ase segments co-ordinated in a single 
sentence have some identical parts not ne- 
cessarily 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~La- 
tire segments are XQ Z and X* - Y*. In ~Lus- 
sian writing ellipsi~ is often accOmpanied 
with a dash. 
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 go- 
ing together° 
BASIC ASSUIAI~TIONS. We examined a co-or- 
dinative ellipsis asslm~in~ the ±'ollowinc: 
A context confined to a standalome el~ 
liptic sentence is sufficient for resto-- 
r~tion. 
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-be- 
ments cover them, too. 
Omitted parts, after restoration :~id, 
if necessary, morphological re-agreement, 
exactly restore the original mea3zing of a 
given sentence as a whole° Therefore recur- 
rent parts in different segments have had 
exactly the soJae length, linear order, and 
lexical content before omission. 
- 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/. 
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. 
There exists an algorithm for syntac- 
tic 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 
65 
sentence and tie links, even a~bigucus, 
within word groul~ in incomplete segments 
preparing them as dis jolted "bushes" for a 
final parse tree° 
STATISTIC OBSERVATIONS° 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). 
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° 
Available statistics permits to cone= 
lude. 
- 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. 
- Cases with N~> 2 and/or NI > 2 amount 
for less than two per cent. . 
Approximately in 7% of cases nonsym- 
metric ("chiastie") formulae occurred° 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@¢ 
~mi~m{o- 
pa )X(pacm~peH~ Ma~oMaT~ec~o~o Epy~oso- 
pa )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¢ 
(the possession of some technical back- 
ground)y, ./ 
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~:. 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 
66 
pairs above are divided with slashos~ ~A 
connected nov_us (derived fz~om complst~ seg~- 
merits) 6u~e given in pa~:enthese~o 
We 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£MN:~ 'qMoJio ~'":-~ !~MO:SI~!/fZ:C H 
Several ~S constitute the :Li~k~~ 'bo~bv~ee~)~ 
a dominating word aud its indi~:oo'Sl~' sdbo:~b- 
dinated one~ Via ES the no(£es ~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° 
The 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:cb 
of 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°/ 
Semantic° 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¢ 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:¢ :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.;:£~7' ,:i"eom ~ ) a{\].{,~tjs.<Lls\]:l)!t 
j,f\] C,>ll'hr?,y ;\],,)_,t {i i O~ i. ~i}~.OJi&t(f~;g(_~ \[J~Z~lD~,>g9 fJO:,.~'~<0-{3¢ 
b~,' ,<!je:~ri;o>x:~co 0;P',~.:,f\].ag :'Ls s~¢o.\].ba:o.eca:u::d.3< do. 
-~,~i'9"@t\]. 3!'370;~4 "{;\]¢.\] 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£t 8"0;ei*<$1;LXX,O 00;~\]£\[ ~)>(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.\[J,{"O72:~_/b\]£.~\] ',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\[;hl :,~,+0 c<>t{'£UT!,O'{;tD~!i~, t6bc;{y ;~\]~O'tfl.(\]. 'bO 
:{.~waz;dL?f ).:L:,F\[,.:,_;d. v!x~. oo,,v'v:,r£u>.ij oo./;.,s J.< <,.<:-. 
G@GI3>-}.\]:,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£1PA,, 
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(~,~} ~'~}£).¢ ; 
\]JO\].'Ti s, ~_tY!t't. i;\]~.Ot~\].({ be ~x~d;J.:o:ood. s.'i; ;Td,~7o:u 'b:,z~;o 
}L(;'¢e\], e, J,J7 ;,9~,g~.oi;:io:~. wo:/v¢.\],<.~ ~.,:¢o OL, tt~otn;rbe:~_.~o(! El.; 
°13I}<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(£t;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¢\]. \[,;%+OT!.\]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\])\[t.obe ¢:m.Oo \]2_o...tt{4~:rteome~{; oJ:' Olin :,.°<~sb~ 
rue.f!, wo'.vSs J.s o~z:¢/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;¢d:i.L~z,,~ 
J;4.'C,'~ 130ft;l~tO))/\[;f\] JZfL 'IU\]~O LIO\]r.d.io:tl.oO 9 j;.\]xO):/ £gi;@\]\]fJ 
} \[JO !j ,'::it.'() :,20\])6&JD"@,{\[ :\[?OZ= (\]'9e&#,~" S'ctbrJeq,tte:rti; 
SOp.;m(;:,:'ti7 <>p'pl;y.Lxleq l):¢eoedltxl 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°e 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.:!: ° 
\]£o(:?.~ 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,~:)'~), ¢.?\],\[l,J,\]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 ;¢ 
vsR:<eT< ~ Z0 o 4.1, ea.~;<, 
)/; td./V;~}.%© 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°J.(~ ,~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£O:i!ii!{S,'{;:LO\]t ;til f-J,ygllll,()'t;.Z'ZC,. CO~2{h\]Yi~,Ob:!.O\]\].L¢~ 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:¢m,:l\]lae 
............. :~-% ................................................................................................... : ............................ 77 ....... { ..... 7--, ...... ; ..... \] .......... 
.: . m 
,,~ "}{ 0 "7):' ::<t t 61;7 \[:e,.,\]X(?oY: ~" ~. w, 
:<-: \] .... ~., 
2 8°0 \[~:,o\]~:0,o~q,~ ~,s :~¢"q,o~":'t)q 
~:> \[b:,!o:~xq,<~yo i,:<,, . .. ~ -<,:,..., v,-,_. 
\[~!,o\[~.,%o:~< 8~ :e t .... :~!:>' 
4, 3<>'? ICP 'iO EO..X' I,'.,~ '~ ,:,c z ~: e.o, C + "!~fl %) kq:.:.~ 
. "O 
d~;/z, om)(~ (o,s, pm~,at,:ox~,:cco!~),< (d!0& 
( l.<\[)Vl (',:K\]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 )-£,, 
( \]'IOZ\]TI b)0b1IIe\])a¶\['yIJ~'~ 0;d~iIS, II;0\]}8 
(:~ -_1:98£ :,~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¢o)- 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)~:+ ,, 
(:<f 't;ho l;o_~i.pe:eo.b~.~:~?o. J.s 
I_.~.@D:~'i.(",&\]. "lT\]:!.e:fl~).,~(i)\].g:tFjNa j.,,-~ • J~ O- 
n~r.o4) Qo (iso~ e:c~8.o):£ (a<uT.)& 
(J.Z 07~\])OSJ.'I;O oase))?q<-, 
(no:oJ.s o-- Ise~,m:Lo):~o 
• o "~r Oo , .. . ...... e~ .J. (O:t" c#o:t.?ea:%;!/uT). Z {.!.:Lo:<'u:~:o.e~ 
~(m~ru:,~-O J:o:r,),, ' (600,~ 0()0 vo\]. o )v ... 
. t~30 . , , (a:n.a)~S (° l: <.Oe.e,:i.a !.£ ~c:d.):£. •, 
• (tZ~?lT,:eo:~dm,.~ge\].~g ~I0 ~ 000 'v'o:k <, )X ':' 
47 
