HIGH-PROBABILITY SYNTACTIC LINKS 
Leonid Mitjushin 
Institute for Problems of Information Transmission 
Russian Academy of Sciences 
19 Ermolovoy Street, 101447 GSP-4, Moscow, Russia 
1 Introduction 
In this paper we consider syntactic relations between 
words of a sentence that can be strongly predicted by 
local mechanisms. For instance, if a sentence con- 
tains a pair of words 
... red block .... 
then the reader immediately makes a conjecture that 
red is an adjective modifier for the noun block. The 
same is true for semantically abnormal pairs such as 
... green ideas .... 
Other examples of strong prediction are provided 
by pairs 
... authors describe .... 
... problem is..., 
for which a "subject - verb" relation takes place with 
high probability. 
In most cases, such simple hypotheses prove to be 
correct. However, sometimes they lead to errors, as 
for the pair problem is in the sentence 
(I) The solution of this problem is very simple. 
In this example, however, by the moment the word is 
has been read, the word problem is already engaged 
in other strongly predicted constructions, namely the 
prepositional phrase of" this problem and even the 
whole noun phrase the solution of this problem. A 
conflict arises, and plausibility of the new hypothesis 
becomes much lower. 
Such syntactic relations may concern not only ad- 
jacent words. For instance, in (1) it is for the pair 
solution ... is that the "subject - verb" relation will be 
conjectured. 
In this paper, slrong prediction of syntactic rela- 
tions is modeled within the framework of dependency 
syntax (see Mel'~uk 1974, 1988). According to this 
theory, (surface) syntactic structure of a sentence is 
an oriented tree whose nodes are the words of the 
sentence (more precisely, their lexico-morphological 
interpretations). The arcs of the tree represent syn- 
tactic links between words and are labeled by names 
of syntactic relations. The result of strong prediction 
is a partial parse of the sentence, in which high-prob- 
ability syntactic links are established. 
In our opinion, dependency structures are better 
adapted to partial parsing than constituent struc- 
tures. The reason is that the dependency structure of 
a segment is the same both when the segment is con- 
sidered as isolated and when it is considered as a part 
of some sentence (by "segment" we understand any 
sequence of words). Generally, this is not true for 
constituent structures. For instance, the segment l 
saw a man has the dependency structure * 
(2) l-compl 
pred \[ det 1 
I ~ SOW O ~ man 
both as a complete sentence and as a part of the 
sentence I saw a man with a telescope. The fact that 
the latter sentence is ambiguous does not hamper 
anything, as both its structures contain subtree (2) 
(and differ only in arcs that go into the word with): 
(3) l-comp\[ prep 
pred \[ del\] attr l, det l 
I ~ saw a ~man ~wi~n a ~ telescope. 
(4) nov 
pred 
I ~saw 
l-compl \[~_ prep 
d¢t \] dnt \[ 
a ~man with a ~ telescope. 
On the other hand, the constituent structure of the 
segment I saw a man is not fully inherited in the 
constituent structures of the longer sentence. In our 
opinion, this comparison demonstrates that, in a cer- 
tain sense, dependency structures reflect the in- 
cremental nature of sentence comprehension from 
left to right better than constituent structures do. 
In this paper we describe a bottom-up, left-to-right 
algorithm of partial parsing that establishes high- 
probability syntactic links. It is implemented on a 
VAX 11/750 computer as a subsystem of a multipur- 
pose linguistic processor developed in the Laboratory 
of Computational Linguistics of the Institute for 
Problems of Information Transmission, the Russian 
Academy of Sciences (Apresjan et al. 1992). The par- 
tial parser is employed as a preprocessing unit before 
the operation of the main filter-type parser. It can also 
be used for automatic indexing and lemmatization. 
The algorithm is language-independent: all lan- 
guage-specific information is recorded in the dic- 
tionaries and the rules that establish links. 
* Full names of English syntactic relationS that appear in example= are: predicative, determinative, lsl completive, prepositional, at- 
tributive, adverbial. The number of relations used In complete models of English and Ru~tan syntax varies from 40 to 
55 (Mel'~uk 
1974; Mel'~.uk and Pertsov 1987; Apresjan et al. 1989, 1992). 
AcrEs DE COLING-92, NANT~, 23-28 AOtJT 1992 9 3 0 PROC. OF COLING-92. NANTES, AUG. 23-28. 1992 
Experiments with Russian sentences have given 
promising results: on average, the algorithm estab- 
lishes 70 - 80 ~o of syntactic links of a sentence; 
processing speed (exclusive of morphological 
analysis) is about 10 words per CPU second. The 
error rate is less than 1% (stable estimates have not 
been obtained yet). 
2 Bottom-up Parsing 
The processing of a sentence begins with morphologi- 
cal analysis. As a result, each word is given a set of its 
possible lexico-morphological interpretations, hence- 
forth called "homonyms". A homonym is a list that 
includes a lexeme identifier, a part-of-speech 
marker, and morphological features of the wordform. 
For instance, the morphological module of the 
ETAP-2 system (Apresjan et al. 1989) will give for 
the word sawtbe following three homonyms: SEE, V, 
pt (-past tense); SAWl, V, mf (-main form); SAW2, 
N, sg. 
All morphological data are concentrated in a spe- 
cial morphological dictionary. The key role in parsing 
proper is played by a combinatorial (syntactic) dic- 
tionary that contains versatile information on syntac- 
tic properties of lexemcs, i.e. on their ability to 
participate in various syntactic constructions (for 
details see Mel'~uk 19"/4, 1988; Apresjan et al. 1989, 
1992). 
The general scheme of parsing is as follows. After 
the morphological analysis, for each word there ap- 
pears one or more homonyms. By "fragment" we shall 
understand a set of homonyms occupying one or 
more successive positions in the sentence (one 
homonym in each position) plus a tree of syntactic 
links defined on these homonyms as nodes. For in- 
stance, an isolated homonym is a trivial fragment; the 
whole dependency tree of a sentence is also a frag- 
ment. It should be noted that in trees (2) - (4) each 
word is represented by a certain homonym (for ex- 
ample, saw is represented by SEE, V, pt). 
Lejkina and Tsejtin (1975) described a bottom-up 
process for constructing dependency trees. It is based 
on the operation of adjunction. This operation is ap- 
plied to two adjacent fragments and consists in estab- 
lishing a link, marked by a certain syntactic relation, 
from a certain node of one fragment to the root of the 
other. The result of adjunction is a new fragment on 
the union of segments occupied by the initial frag- 
ments. 
This action is similar to generation of a new con- 
stituent from two adjacent constituents. However, 
unlike constituents, fragments at the moment of ad- 
junction may be "incomplete", i.e. they need not con- 
tain all the nodes that will be direct or indirect 
dependents of their roots in the structure of the sen- 
tence. These nodes may be added to them later (also 
by the operation of adjunction). 
Mitjushin (1985) described the class of trees that 
can be constructed from isolated homonyms by 
repeated adjunction, i.e. that can be built by the bot- 
tom-up process. Consider a tree with an ordered set 
of nodes. Let a "tangle" be a quadruple of nodes (a, b, 
c, d) with the following properties: 
1) a<b<c<d; 
2) a and c are linked by an arc (in any direction); 
3) b and d are linked by an arc (in any direction); 
4) the path between a and d contains neither b nor 
c (here, orientation of arcs is ignored, so the 
path always exists). 
The following criterion is true: a tree can be con- 
structed from its nodes by repeated adjunction if and 
only if it contains no tangles. 
The simplest tangle looks as follows: 
a b c d 
(direction of the arcs does not matter; there can be 
other nodes between a, b, c, and d). According to the 
criterion, a tree that contains such a subtree cannot 
be built by the bottom-up process. 
The class of trees obtainable by adjunction is much 
wider than the class of so-called projective trees (on 
projectivity see, for example, Gladkij 1985; Mel't~uk 
1988). For the model of Russian syntax presented by 
Mel'~uk (1974) and Apresjan et al. (1992), this class 
includes all syntactic structures permissible in scien- 
tific, technical, and business texts (however, it is not 
so for spoken language and poetry). We suppose all 
the structures considered below to belong to this 
class. 
3 Rules 
In our system, in contrast to those based on forulal 
grammars, the rules are not a tool for the exhaustive 
description of the set of correct syntactic structures. 
We suppose that the correspondence between senten- 
ces and their syntactic structures is defined by some 
other means. The task of the parsing algorithm and 
the rules it employs is to build, for a given sentence, 
some set of its syntactic structures or their fragments, 
without losing the semantically correct ones. 
The concrete function of the rules is to check 
whether the given case of adjunction is feasible and, 
if so, to perform the operation of adjunctioa. Some 
additional operations can also be performed. The 
rules have access to any information about the struc- 
ture of fragments to be adjoined and the homonyms 
they contain (their lexeme names, morphological fea- 
tures, and syntactic properties stated in the com- 
binatorial dictionary). The rules may also use data 
on punctuation and limited data on homonyms not 
belonging to the given two fragments; they have no 
access to information about fragments built by the al- 
gorithm earlier. 
While formally the rules could be strongly context- 
sensitive within the limits of two given fragments, in 
most cases they only use information on nodes X and 
Y (those to be linked) and their nearest syntactic 
Context. In fact, the rules currently emloyed do not 
ACTES DE COLING-92, NANTES. 23-28 Am~'r 1992 9 3 1 PROC. OF COLING-92, NANTES, AUG. 23-28. 1992 
consider nodes for which distance from X or Y ex- 
ceeds 3 (where distance is the number of links in the 
path connecting two nodes in the dependency tree of 
a fragment). 
A rule is a program written in the form of a transi- 
lion graph, with an elementary predicate or operation 
associated with each arc. The rule interpreter per- 
forms ordered search to find a path along "true" arcs 
that starts at a fixed entry node and ends at one of 
fixed exit nodes. No backtracking is used: if forward 
motion from some node proves to be impossible, in- 
terpretation is terminated. The fact that backtracking 
is not necessary has been discovered empirically; it is 
connected with the nature of syntactic events con- 
sidered by the rules. On the other hand, when 
desirable, an explicit return may be made to a point 
passed earlier, with simple measures taken against 
infinite cycling. 
Each rule contains at least one operation 
LINK(X, Y,R) that establishes a link marked by a 
certain syntactic relation R between the given node X 
of one fragment and the root Y of the other (that is, 
performs the adjunction). The corpus of rules covers 
only those situations for which the probability that 
the established links are correct is estimated as close 
to 1. For instance, the rules do not establish links like 
alter and ad~v in structures (3) and (4) because altach- 
ment of prepositional postmodifiers is known as a 
"classical" case of structural ambiguity. 
It should be noted that the probability close to 1 
characterizes here not individual links (it would be 
too strong a demand) but all complex of links estab- 
lished for the given words. This can be illustrated by 
the segment 1 saw, for which two fragments will be 
built with different homonyms for the word saw: 
pred pred 
I~-- SEE V.pt, I~ SAWI V,mf. 
Both these alternatives are syntactically correct. At 
the same time, they are mutually exclusive, and it is 
only their disjunction that has probability close to 1. 
This ambiguity is also inherited by larger fragments. 
(As a result, the sentence 1 saw a man with a telescope 
has four different parses, two of which are semanti- 
cally abnormal.) Thus, high probability is a "roller- 
five" and not an "individual" property of links. 
Rigorous definitions can be found in the paper by 
Miljushin (1988). 
4 The Parsing Algorithm 
The simplest method of bottom-up parsing is to con- 
sider all opportunities for adjunction, starting from 
adjacent one-element fragments. We employ a faster 
algorithm, in which certain heuristics are used to 
reduce search (Mitjushin 1988). 
The algorithm builds a growing sequence A of frag- 
ments. At any moment of time A contains some 
homonyms of the sentence and certain fragments 
constructed of these homonyms. The algorithm 
moves from the beginning of the sequence A to its end 
and tries to perform adjunction between the current 
fragment F E A and the fragments that appear in A 
earlier than F. New fragments are added to the end of 
the sequence. 
The fragment cmtsidered at the given moment is 
called active. All fragments of A (including isolated 
homonyms) become active successively, without 
leaps or returns. 
While the algorithm moves along the sequence A, 
tile sequence grows longer because of the addition of 
newly built fragments. Nevertheless, a moment will 
necessarily come when the active fragment is the last 
in A and further motion is impossible. In this case, 
the next homonym of the sentence is added to the se- 
quence; it becomes active and the work is continued. 
When a new deadlock arises, another homonym is 
added, and so on. If in such a situation it turns out 
that all homonyms of the sentence are exhausted, 
then the work is finished. 
Homonyms are added to the sequence in the order 
they are arranged in the sentence from left to right 
(which is essential), and those occupying the same 
position are added in an arbitrary order (in this case, 
the order has no influence m~ the results). At the ini- 
tial moment A contains a single element, namely one 
of the homonyms occupying the leftmost position of 
the sentence, and it is declared active. 
For each active fragment Flhe algorithm selects in 
A its left neighbors, i.e. fragments thai are adjacent 
to F on its left. A preference relation is defined be- 
tween tile neighbors of F: fragments of greatter length 
are preferred, and lhose of equal length are con- 
sidered equivalent. 
For the given F, the algorithm considers its left 
neighbors E in order of their preference, and for each 
E tries to adjoin it to F. If for some E adjunclion is 
successful, subsequent search is limited to the neigh- 
bors of F equivalent to E; less preferred fragments are 
not considered. 
An attempt to adjoin E to F is made as follows. 
IJnks are considered that connect a certain node X of 
fragment E with the rightmost node Y of fragment F. 
A preference relation is defined between the links: 
those of greater length are less prefen'ed, and those 
of equal length are equivalent. In other words, more 
preferred are links X -- -Y and X ~ Y with nodes X 
that are nearer to the Jight end of E; links with the 
same X are equivalent. 
For the given E and F, nodes X ~E are considered 
from right to left (i.e. in order of the preference of 
links between X and Y), and for each X the rules ap- 
plicable to these X and Y are activated. The list of 
such rules is determined by parts of speech of X and 
Y, and by possible direction of fhe link. If during in- 
terpretation of a rule an operation LINK(X, Y, .) or 
LINK(Y,X, .) is performed then a new fragment is 
built which is the ~'esult of joining X and Y with the 
given link. It is placed at the end of tile sequence A. 
After flint, for these E and F the search is limited to 
the links equivalent to the established one; less 
preferred links are not considered. 
When the sequence A is built, its subset C of maxi- 
mal fragments is formed. A fragment is called maxi- 
mal if its segment is not a proper part of the segment 
of any other fragment belonging to A. The set C is the 
final result of partial parsing. Below, when speaking 
ACTES DE COLING-92, NANTES, 23-28 AO(ZI' 1992 9 3 2 PROC. OI' COLING 92, NANI KS. AUG. 23-28. 1992 
about fragments built by the algoritlnn, we shall al.- 
ways mean exactly tim set C. 
The first experiments with this algorithm have 
shown titbit, tit sonte c;.tses, the preferences attd 
restrictions adopted arc too strong and pvtme away 
semantically correct parses. To intprove tire situa- 
tion, special operations were defined that made it 
possible to cancel (from inside lhc rule) priority of 
longer neighbors or shorter links, and also to make 
the algm'ithm cmrsider not only the rightmost node of 
tire right fragment. ()wing to them, the search can be 
made exhaustive in all cases when the rule "considers 
it desirable". In tile real process of pursing, these 
operations are fired not too often, so tile main part of 
search remains limited. 
5 Experimcn|s 
At present, after preliminary debugging and tuning of 
tile rules, we have begun to carry out regahn' experi~ 
merits with it homogeneous flow of Russian texts. The 
experiments make use of a Coluputer-olicnted conl- 
binatorial dictionary of Russian compiled by a group 
of linguists under ttle guidance of Ju.D.Apresjan (see 
Apresjan et al. 1992). It contains over' 10,000 entries, 
mainly general scicnlific vcxzabulary and terms horn 
computer science and e\]tx:trical engineering. 
The number of rules in lhc system is now about 
100. Total number of arcs in their transition graphs is 
about 2,000. 
As a source of texts, we have taken several issues of 
the journal Computer Science Abstracts (Referativnyj 
zhurnal Vyehislitel'nyje Nauki, in Russian). Senten- 
ces are chosen at raodom. Sentences with formulas, 
occasional abbreviations, and non-Cyrillic words are 
excluded. Words absent in the dictionaries (aboul 
8% of all word occureuces in these texts) are 
replaced by "dummy" words that have syntactic 
properties most probable for the given category. At 
present, about 300 sentences have been processed. 
On the average, fraginr:nts produced by partial 
parsing include 3 - 4 words. It is not infrequent that 
they have 8 - 10 or store words, or present complete 
structures of sentences. On the other hand, a sub- 
stantial parl of fragments are isolated homonyms. 
For instance, subordinate conjunctions remain iso- 
lated in most eases because, as a rule, their links wilh 
other words are not considered having high prob- 
ability. 
Frequently enough morphoh~gieal, lexical, and 
structural ambiguity results ill building 2 - 4 different 
fragments on tile same segnlellt, Sometimes their 
number is 8 - 12 and more, but such cases are rela- 
tively rare. The record is now equal to 72 fragments 
on a segment of 9 words. For such cases, packing 
techniques can be developed similar to those 
described by Tomita (1987). Another possible 
method is to employ ntnnelical estimates of syntactic 
preference (set, for example, Tsejtin 1975; Kulagiua 
1987, 1990; Tsujii et al. 1988). 
On the avecage, the nmubcr of established links is 
70 - 80 % of the total nunlber of syntactic links in tile 
sentence. These figm'es include links present both in 
the fragmenls built ;trl0 ill tile semantically COl r¢ct 
structm'e of the sentence; "extra" links that arise due 
to ambiguity of fragments are not included. 
Sometimes the fragments overlap, that is, their 
segments intersect. It happens approximately in one 
tenth of sentences. As a rule, in such cases the correct 
resnlt is a combination of one of the overlapping frag- 
ments with its "truncated" competitor. 
A fragment is called correct for a given sentence if 
it is a subtree of the semantically correct dependency 
trek of this sentence (or of one of such trees, in the 
rare cases of real semantic ambiguity like (3) - (4)). 
A h'agment is called feasible if it is a subtree of some 
dependency tree of some sentence of the given lan- 
guage. The algmSthm makes an error in the following 
lwo cases: (a) if a non-feasible fragment is built; (b) 
if all fragments built on some segment are feasible 
but none is correct. (Here we do not take into account 
semantically abnormal sentences or the possibility of 
overlapping; these situations would require more ac- 
curate definitions.) 
hi roost cases, all error means that some link of a 
fragment is established erroneously, while all the 
others arc correct. Ttre experiments have shown that 
tile frequency of errors for the algorithm described is 
fairly snmll. For tile lasl 100 sentences, 12 errors 
were nmde (9 of the first type and 3 of the second), 
which is less than 1% of the total number of links 
eslablished in correct fragments. A stable estimate is 
not yet obtained because at this stage of experiments 
tuning of tire rules is emllinued, and the error fre- 
queocy decreases steadily. 
Error s of tire first type are caused by inaccuracy of 
the lexicographic descriplious and imperfection of 
the rules. In the presence of adequate lexicographie 
information, these errors in principle are avoidable, 
as the rules may fully control internal properties of 
the fragments being created. 
The second type of error is intrinsic to our ap- 
proach. The rules employed are local in two respects: 
they take no (or almost no) account of the context 
outside the fragments being adjoined, and they take 
no account of a very large part of syntax that concerns 
less probable links. The first restrictiou means that 
fragments may appear which are grammatically 
feasible but do not agree with the context. The 
second one implies that wc do not intend to obtain 
complete structures of sentences, and therefore shall 
no\[ be able to reject a fragment for the reason that it 
is not engaged in any complete structure. 
In general, it is not at all snrprising that a certain 
part of syntactic links can be reliably revealed by 
local mcchanisrns. Any flow of texts in any language 
must contain chains of words the parse of which 
weakly depends on the context ("weakly" can be un- 
derstood here in the statistical sense: the share of 
those occurences for which tile parse differs from the 
most probable one is small). The possibility of ex- 
amining fragments in any detail permits to avoid 
situations iu which the risk of creating a non-feasible 
fragment is too large. 
A more surprising fact is that the number of reliab- 
ly established links is rather high ~ about 75 %. For 
the most part, these are links typical of the basic, 
most frequent syntactic constructions such as "adjec- 
ACIES DI,;COLlN(l-92, NANIV;S, 23-28 AOt-ll 1992 93 3 l'aoC. OV COLING-92, NAhqES, AUG, 23-28, 1992 
tire + noun", "preposition + noun", "numeral + noun", 
"adverb + verb", and also a large group of links con- 
necting predicate words with their arguments. As 
regards the last type, preference for the predicate-ar- 
gument interpretation of word combinations was 
orlon noted in the literature (this preference is a par- 
ticular case of the Most Restrictive Context Principle 
proposed by Hobhs and Bear (1990)). 
Observations show that the number of established 
high-probability links noticeably depends on the type 
of text. The general trend is as follows: the more "for- 
mal" the text is, the more links are established. From 
this point of view, the language of scientific abstracts 
suits the given approach quite well. 
As regards comparative frequency of high-prob- 
ability links in different languages, it would be 
natural to expect these links to be more typical of lan- 
guages with rich morphology than of analytical ones 
(such as English). Nevertheless, preliminary experi- 
ments have shown no substantial difference in this 
respect between English and Russian scientific texts. 
We suppose that in case of high-probability links, 
the efficiency of local approach is additionally aug- 
mented due to factors "of the second order" concern- 
ing general mechanisms of text comprehension and 
generation. This opinion is based on the following as- 
sumptions. If someone reading a text sees that a 
high-probability link is possible between certain 
words and this link is compatible with the previous 
part of the text, then he makes a conjecture that this 
link is correct; such conjecture is abandoned only if 
some counter-evidence is obtained. When people 
generate texts, they take into account this property of 
the comprehension mechanism and tend not to disap- 
point expectations of the readers. In other words, 
they are careful not to create high-probability links 
that would prove to be incorrect. This can be re- 
garded as an instance of cooperation in language per- 
formance (cf. the Cooperative Principle in pragmatics 
formulated by Grice (1975)). 
References 
Apresjan, Ju.D., I.M.Boguslavskij, L.L.Iomdin, 
A.V.Lazurskij, N.V.Pertsov, V.Z.Sannikov, and 
L.L.Tsinman. 1989. Lingvisticheskoje Obespe- 
chenije Sistemy ETAP-2. Nauka, Moscow. ('The 
linguistics of the ETAP-2 system', in Russian) 
Apresjan, Ju.D., I.M.Boguslavskij, L.L.Iomdin, 
A.V.Lazurskij, LG.Mitjushin, V.Z.Sannikov, and 
L.LTsinman. 1992 (forthcoming). Lingvisticheskij 
Protsessor dlja Slozhnykh Informatsionnykh Sis- 
tern. Nauka, Moscow. ('A linguistic processor for 
complex information systems', in Russian) 
Gladkij, A.V. 1985. Sintaksicheskije Struktury 
Estestvennogo Yazyka v Avtomatizirovannykh Sis- 
temakh Obshchenija. Nauka, Moscow. ('Syntactic 
structures of natural language in automatic 
dialogue systems', in Russian) 
Grice, H.P. 1975. Logic and Conversation. In P.Cole, 
J.L.Morgan, editors. Syntax and Semantics, Vol. 
3, Academic Press, New York, pp. 41 - 58. 
Hobbs, .LR. and J.Bear. 1990. Two Principles of 
Parse Preference. In Proceedings of COLING-90, 
Vol. 3, Helsinki, pp. 162 - 167. 
Kulagina, O.S. 1987. Ob Avtomaticheskom sintak- 
sicheskom Analize Russkikh Tekstov. Preprint No. 
205, Institute for Applied Mathematics, Moscow. 
('On automatic parsing of Russian texts', in Rus- 
sian) 
Kulagina, O.S. 1990. O Sintaksicheskom Analize na 
Osnove Predpochtenij. Preprint No. 3, Institute for 
Applied Mathematics, Moscow. ('On preference- 
based parsing', in Russian) 
Lejkina, B.M. and G.S.Tsejtin. 1975. Sintaksiches- 
kaja Model' s Dopushchenijem Ogranichennoj 
Neprojectivnosti. In Meztulunarodnyj Seminar po 
Mashinnomu Perevodu, Moscow, pp. 72 - 74. ('A 
syntactic model allowing limited non-projectivity', 
in Russian) 
Mel'~uk, I.A. 1974. Opyt Teorii Lingvisticheskikh 
Modelej "Smysl ,,* Tekst". Nauka, Moscow. 
('Toward a theory of Meaning ~--, Text linguistic 
models', in Russian) 
Mel'tfuk, I.A. 1988. Dependency Syntax: Theory and 
Practice. State University of New York Press, Al- 
bany. 
Mel'~uk, I.A. and N.V.Pertsov. 1987. Surface Syntax 
of English." A Formal Model within the Meaning 
--- Text Framework. John Benjamins, Amster- 
dam. 
Mitjushin, L.G. 1985. Dlina Sintaksicheskikh Svjazej 
i Induklivnyje Struktury. In Semiotika i Infor- 
matika, No. 26, Moscow, pp. 34 - 51. ('Length of 
syntactic links and the class of inductive 
structures', in Russian) 
Mitjushin, LG. 1988. O Vysokoverojatnykh Sintak- 
sicheskikh Svjazjakh. In Problemy Razrabotki 
Formal'noj Modeli Jazyka (series "Voprosy 
Kibernetiki", No. 137), Moscow, pp. 145 - 174. 
('On high-probability syntactic links', in Russian) 
Tomita, M. 1987. An Efficient Augmented-Context- 
Free Parsing Algorithm. Computational Linguis- 
tics, Vol. 13, No. 1 - 2, pp. 31 - 46. 
Tsejtin, G.S. 1975. Metody Sintaksicheskogo 
Analiza, Ispol'zujushchije Predpochtenije 
Jazykovykh Konstruktsij: Modeli i Eksperimenty. 
In Mezhdunarodnyj Seminar po Mashinnomu 
Perevodu, Moscow, pp. 131 - 133. ('Parsing 
methods based on preference of the language con- 
structions: models and experiments', in Russian) 
Tsujii, J., Y.Muto, Y.Ikeda, and M.Nagao. 1988. 
How to Get Preferred Readings in Natural Lan- 
guage Analysis. In Proceedings of COLING-88, 
Vol. 2, Budapest, pp. 683 - 687. 
ACRES DE COLING-92, NANTES, 23-28 ^OI~T 1992 9 3 4 PROC. OF COLING-92, NANTES, AUO. 23-28, 1992 
