Anaphor Resolution and the Scope of Syntactic Constraints 
Roland Stuckardt 
German National Research Center for Information Technology (GMD) 
K()NTEXT - Natural Language Systems 
Dolivostrafie 15, D-64293 Darmstadt, Germany 
stuckar(l(o)darmst adt .grad.de 
Abstract 
An anal)hor resolution algorithm is pre- 
sented which relies on a combination of 
strategies for narrowing down and select- 
ing ti'om antecedent sets fl)r reflexive pro- 
nouns, nonreflexive pronom~s, and com- 
mon 11011118. ~lqle work focuses on syn- 
tactic restrictions which are derived froin 
Chomsky's Binding Theory. It is dis- 
cussed how these constraints can be in- 
corporated adequately in an anaphor res- 
olution algorithm. Moreover, by showing 
that t)ragmatic inferences may t)e neces- 
sary, the limits of syntactic restrictions 
are ehleidated. 
1 Introduction 
It is by now widely agreed upon that tile process of 
resolving anaphors in natural language text is sup- 
ported by a w~riety of strategies employing differ- 
ent kinds of knowledge. The t)rocess of determin- 
in9 the set of possible antecedents is governe, d by 
morphosyntactic, syntactic, semantic, and prag- 
Inatic restrictions. The same holds for preferences 
applie, d in the antecedent selection process: simple 
surface criteria are involved as well as more elab- 
orate syntactic, semantic, or focusing heuristics. 
As a consequence, recent approaches to anaphor 
resolution apply a careflflly selected blend of con- 
straints and preferences, thus constituting Inulti- 
strategy approaches in the sense of Carbonell and 
Brown (Carbonell and Brown, 1988). 
There are, however, implementability limita- 
tions. At discourse level, determining the set of 
admissible antecedents requires a representation 
which is ordered according to pragmatic relations 
(Grosz and Sidner, 1986; Wehber, 1989). Al- 
though various theoretical frmneworks have been 
suggested, the recognition of these relations in 
the case of unrestricted discourse is still beyond 
the state-of the-art. Moreover, there arc cases ill 
which antecedent decisions can only be made on 
the grounds of domain knowledge and inferencing, 
and although there have been various attempts to 
integrate components of these kinds into anaphor 
resolution approaches, a satisfying solution l;o this 
problem is not available by now. 
As a conseqllence, c/lrrent anaphor resolution 
implementations rely oil constraints and prefer- 
ence heuristics which employ information origi- 
nating from morphosyntactic, syntactic, or shal- 
low semmltic analysis (of. (Carter, 1987)). These 
approaches, however, perform remarkably well. 
An early case study revealed that a 'naive' algo- 
rithm for resolving nonre\[texive pronouns, by re- 
lying merely on morphosyntactic, syntactic, and 
surface criteria, yields correct results for more 
than 80 percent of pronoun occurrences, and that 
tile incorporation of selectional constraints re- 
suits in a gain of another 3.5 percent of accuracy 
(Hobbs, 1978). These results have been confirmed 
by recent work (Lappin and Leass, 1994). The lat- 
ter approach is based on a more elaborate, theory- 
oriented, declarative, forlnulation of the syntactic 
constraints, and handles reflexive pronouns too. It 
proved that the incorporation of statistically mea- 
sured lexical preference patterns (a dynamic, do- 
main specific suhstitute R)r the static encoding of 
sele('tional t)references) yields a gain of only 3 pe.r 
cent, and a sol(; application of lexical preference 
t)atterns resulted in a performance below 35 per 
cent. Hence, there is strong evidence that syntac- 
tic restrictions in combination with surface based 
and syntactic prefe, rence criteria play the central 
role in realistic approaches to anaphor resolution. 
In this I)aper, an anaI)hor resolution algorithin 
is described which has t)een implemented as part 
of the KONTEXT text analysis system for the 
German language (Haenelt, 1994). The empha- 
sis lies on the description of imt)lementation tech- 
niques for syntactic constraints. Section 2 works 
out strategies whi(:h are applied, focusing on the 
937 
theoretical background from which the syntactic 
constraints emerge. Section 3 describes how these 
strategies are coined into an algorithm for the res- 
olution of reflexives, nonreflexive pronouns, and 
definite common nouns, thereby elucidating de- 
tails which have to be taken into account in an ad- 
equate implenmntation. In section 4, a theoretical 
evaluation is performed, and application results 
are given. Section 5 points out that the structural 
constraints may depend on circumstances which 
are not a matter of syntax alone, but rather ne- 
cessitate semantic and pragmatic infereneing. As 
a consequence, limitations concerning the imple- 
mentability show up, and the scope of syntactic 
constraints proves to be restricted. 
2 Constraints and Preferences 
2.1 Morphosyntactic Agreement 
A quite strict constraint requires the pronoun to 
agree with its antecedent in person, number, and 
gender. In example 1 
(1) The father visited his daughter. 
She had invited him on Sunday. 
the antecedents for hint and size are identified 
uniquely as father and daughter', respectively. 
2.2 Syntactic Constraints 
The following data substantiate the syntactic re- 
strictions which are to be employed: 
(2a) The barbe~ shaves hirnselfi. 
(2b) * The clienti appreciates 
that the barbcr shaves himselfi. 
These examples suggest that reflexive pronouns 
choose their antecedents in some kind of local do- 
main. On the other hand, examples 
(3a) * The barberi shaves himi. 
(3b) The elienti appreciates 
that the barber shaves himi. 
indicate that the admissible structural positions 
of antecedents for nonretiexive pronouns are dis- 
tributed complementarily, i.e. these pronouns 
choose their antecedents outside of their local do- 
main. An even more stringent restriction holds 
for nonpronominal nouns: 
(/ta) * The barber) shaves the barberi. 
(~b) * The client,i appreciates 
that the barber sttaves the clicnt,i. 
But even here, configurations exist in which in- 
trasentential antecedents are possible: 
*The examples are given in English. The phenom- 
ena and its implications translate directly to German. 
(4c) The barber who shaved the clienti 
told the elienti a story. 
Ctlomsky provides a formal description of these 
observations as part of his Government and Bind- 
lug (GB) Theory (Chomsky, 1981; Chomsky, 
1986). Binding Theory (BT) distinguishes three 
types of NP, namely type A ('anaphor', compris- 
ing reflexives and reciprocals2), type B (nonreflex- 
ire pronouns), and type C ('referring' expressions, 
comprising common nouns and names). The re- 
strictions are stated as binding principles: 
Definition 1 (binding principles) 
(A) An anaphor is bound in its binding category. 
(B) A pronominal is free (i.e. not bound) in its 
binding category. 
(C) A referring expression is free (i.e. not bound) 
in any domain. 
where binds is a relation which is defined on the 
NP nodes of the (surface) phrase structure tree: 
Definition 2 (the binding relation) Node X 
binds Node Y if and only if X and Y arc coin- 
dexed and X e-commands Y. 
where (definitions vary slightly): 
Definition 3 (the c-command relation) 
Node X e-commands node Y if and only if the next 
b~nnehing node which dominates X also dominates 
Y and neither X dominates Y, Y dominates X nor 
X=Y. 
The central part of the Binding Theory develops 
the notion of local domain to which binding prin- 
ciples A, B, and C refer as binding category: 
Definition 4 (binding category) Node X is 
binding category of node Y if and only if X is the 
next node which dominates Y, and which contains 
a subject that e-commands Y. 
Due to these definitions, the acceptability judge- 
ments for the data presented above are reproduced 
by binding principles A, B, and C. For each exam- 
ple, the subject demarcating the (local) binding 
category is just the ordinary subject of the subor- 
dinate clause. (One has to recall that, in phrase 
structure trees, the subject c-commands the con- 
tent of the VP.) The notion of subject, however, 
is a more general one, applying also to some kinds 
of nominal phrase attributes, in particular certain 
variations of genitives and possessives: 
(,5) Peter listens to Sam'si story about himself. 
2In this paper, the notion of anaphor is used more 
generally. When referring to anaphor in the Chom~ 
skyan sense, the notion reflexive/reciprocal (pronoun) 
is used. 
938 
2.3 Antecedent Predictability 
For eataphorie pronominal resumptions, a con- 
straint is applied which has l)een described l)y 
Kuno (Kuno, 1987). According to 
(6@ The barber who shaved himi 
told the: client| a story. 
(6b) * The barber who shaved him| 
told a clienti a story. 
a dcJiniteness requirement has to be fulfilled, rul- 
ing out antecedents which are not predictable, i.e. 
not a.lready introduc, ed in the. discourse° 
2.4 Case Role Inertia 
in g(meral, the constraint applicatioil will not sin- 
gle out a uifique antecedent. Depending on the 
tyl)e of anaphor to be resolved, preferenc(,s are 
applied, coinprising the rather superficial and self- 
exf)lanatory criteria of recen(:y, cataphor penalty, 
and sul)ject preference. The case role inertia cri- 
terion, which proved to/)e very useful in practice, 
is explainal)h; by the following e×amt)le: 
(7) Peter visited his brvther. 
lte showed him his ne'.w car. 
Unless given further information, there see, ms to 
I)e a strong tendency to choose the antecedents 
in a way that the, syntactic and/or semantic case 
roles of the pronouns re, produce the correspond- 
ing roles of the, it antecedents. Thus, the pre, fer- 
ence rule suggests Peter as the, ~mtecedent for lie, 
and brother as the antecedent for him. As can t)e 
demonstrated by fllrth(,'r e, xamples (e.g. changing 
from active, to passive voice or vice versa), retain- 
ing the semanti(: case role should outvote retaining 
the synta(:tic (:as(; role. In cases in whi(:h semantic 
(:ase is not available, however, promoting syntactic 
(:as(', t)arallelism serves as a good at)proximal|on. 
In its effect, this prefl;rence rule al)proximat(;s 
the' often suggested heuristic of ke, eping rather 
then shifting ret?;rential focus (of. (Sidner, 1983)). 
3 Towards the Algorithm 
The lnaill question concerns the adequate imple- 
mentation of Chomsky's I)inding t)rinciples. Some 
a l)riori remarks on theoretic subtM;ies and on the 
eml)loyed ret)resentation are in t)lace. 
3.1 lnterdei)endency Sensitiveness 
As state(l t)y (Correa, 1988), an immediate imI)le- 
men|at|on of th(; constraints proposed in Binding 
Theory is unlbasil)le. Chomsky states, merely as a 
the()rctical device, a flee, inclexing rule wlfich ran- 
(lomly assigns reference in(lexes to surface struc- 
ture NP nodes. During inapt)lug to the seinanti(: 
LF (logical form) representation, the t)inding prin- 
ciples s()~'ve as restrictions tbr filtering out the im 
dex distributions which are considered valid when 
intert)re, ted as eorefL'rence markers. A direct im- 
l)leme, ntation of this generate-~md-test 1)ro(:edure 
yields an exponential time complexity. 
Current approaches avoid gen('rate-and-te~st |)y 
resorting to different strategies. According to 1;11(; 
most colnmon tectmiquc, for anat)horic NPs, a 
separate antecedent search is t)ertbrmed, resulting 
in a quadratic time complexity (e.g. (Hot)bs, 1978; 
Strube and Hahn, 1995)). Because, howev(;r, the 
ante(:edent decisions are performed in isolation, 
invalid index distributions may m'ise. In examph; 
(Sa) The barbcri told the elientj a story, 
while hek shaved himl. 
neither of the t)ronouns is confined structurally to 
one of the intrasenttmtial antece.(lent eandi(tates in 
tie matrix clause. But, afl, era first decision, e.g. 
(Sb) The barberi told the clientj a story, 
while he| shaved himz. 
the situation changes, for one of the antect;dent 
options of the still unresolved l)ronoun is no longer 
available. Binding principle B may bc violat(.'d: 
(8c) * The barberi told the. clientj a story, 
while he| shaved him|. 
An interdepemlen(:y 1)el;we(;it antecedent; choi(:es 
may arise as well when choosing/)etween discourse 
alltece(leiltS~ OF as a COllSeqll(;llc( ~, ()f relative, clause 
attachment, which 1)redetermines coindexing. 
The at)proach presented below is sealsitive to 
these, decision interdependencies, while avoiding 
the exponential time comi)lexity of an immedL 
ate l)inding constraint implem(mtatioil. This is 
achieved by supplementing the straightforward se,- 
quential strategy with a dynamic reveritication of 
the binding restrictions in the antecedent selection 
stet). To avoid that (te.sirable antecedent options 
are ruled out l)y interdependency, the choices wil;h 
highest plausibility is given preference to. 
3.2 Representing Surihce Structure 
The original statement of Binding Theory forms 
part of GB Theory, in which a broader set <)1' in: 
tera<:ting l)rin(:iph~s is f<)rmulated. Because the 
aim of aimi)hor resolution for a specific language 
is restricted, the reI)resentation (:an be simplifiexl. 
Complicating details which result fl'om the Gll 
claim to mfiversality may t)e emil;ted. 
lies|des being efficiently searchable, the simpli- 
fied surface structure has to represent the stru(:- 
t;ural details wtfich are necessary for th(,' verifica-. 
tion of the 1)in(ling restrictions. In particular, this 
939 
comprises subject-object-asymmetry, the demar- 
cation of local domains, and surf.ace order depen- 
dent structural variations 3. 
Because the KONTEXT text analysis system is 
based on a dependency grammar, a mapping pro- 
cess generates the required representation from 
a dependency trees which is not suitable for a 
structural verification of the binding principles, 
because vital details are not structurally visible. 
The attempt of directly Verifying BT restrictions 
on dependency structure, as suggested by Strube 
and Hahn (Strube and Hahn, 1995), does not seem 
adequate, because important details are ignored. 
The structures which were generated for some 
of the above examples are as follows: 4 
(ga) (S barber (VP himself)) -+ (2a) 
(gb) client 
(VP (STHAT barber (VP him)))) -+(3b) 
(9c) (S barber (SREL who (VP client)) 
(VP client (VP story))) -+(4c) 
The marker nodes STHAT and SREL are delim- 
iters of local domains, to which the binding prin- 
ciple verification functions are sensitive. 
Special techniques are employed in representing 
local NP domains, which are introduced by de- 
verbative NPs and NPs with possessive markers 
(saxonian genitive, genitivus possessivus, posses- 
sive pronoun, or certain attributive PPs), e.g. 
(10) The barber hears hisi story about himselfi. 
(S barber 
(VP storyj 
(SVATT x_storyj 
(ATT his (ATT (PP himself)...) 
A domain SVATT enforcing local reflexivation is 
opened. The NP barber and the reflexive pronoun 
himself may be coindexed only indirectly via the 
possessive pronoun his, which is of type B, and 
hence forced to take a nonlocal antecedent. In 
accordance with intuitive judgement, a local in- 
stance of the NP storyj blocks the eoindexing of 
the possessivc pronoun and its dominating noun. 
Here again, the mechanism which copes with in- 
terdependencics is appliedfi Technically, new NP 
types C' (example (10)) and B' (relative pronoun, 
3This concerns certain cases of subject and ob- 
ject clause extraposition as well as, in particular, the 
object NPs contained in the VP, for which a right 
branching structure is generated, yielding a base for 
a structural determination of admissible antecedents 
for reflexive pronouns, which is mainly governed by 
subject-object asymmetry and surface order. 
4Implementation details are ignored. 
5This technique resembles the use of traces in 
Chomsky's GB theory. Because of its restricted aim, 
however, it is nmch simpler. 
cf. section 3.1) are introduced for which binding 
principles C and B are verified, respectively, but 
for which no antecedent search is performed. 
3.3 The Algorithm 
The KONTEXT anaphor resolution algorithm, as 
shown in figure 1, consists of three phases: con- 
straint application, preference criteria application 
and plausibility sorting, and antecedent selection 
including reverification of constraints which may 
be involved in decision interdependencies. 
Two binding constraint verification procedures 
are employed which differ in the handling of type 
A NPs. According to binding principle A, a re- 
flexive pronoun requires 'constructively' a local 
antecedent (step l(b)i). Example (10), however, 
illustrates that further nonloeal coindexings are 
admissible. This gives rise to a weak version of 
binding constraint verification, the usage of which 
is of vital importance to the fimctioning of the 
interdependency test step 3b. 
4 Evaluation 
As a proper base for comparison, the theoreti- 
cal analysis is restricted to the contribution of in- 
trasentential antecedent search. Let n be the num- 
ber of NP nodes in the surface structure represen- 
tation. Because the number of anaphoric NPs and 
intrasentential candidates is bounded by n, and 
the individual a priori verifications of the bind- 
ing principles contribute costs proportional to the 
number of nodes in the surface structure tree, the 
worst case time complexity of step 1 is O(n3). A 
similar analysis, assuming a clever handling which 
prevents individual interdependency checks from 
being done more then once, reveals that the com- 
plexity of step 3 is O(n 3) too. Therefore, since 
the scoring and sorting step 2 does not exceed this 
limit, the overall worst case complexity is O(n3). 
In tests on architect biographies drawn from 
(Lampugnani, 1983), the algorithm correctly re- 
solved approximately 90 per cent of type B pro- 
nouns (including possessives), and, as expected, 
all occurrences of reflexives, which occur quite 
scarcely in the test corpus. The set of possible 
antecedents tends to be reduced drastically during 
constraint application. Interdependency collisions 
did not happen too frequent. This tendency is 
strongly supported by the case role inertia heuris- 
tic, which promotes a complementary distribution 
of preferred antecedents for type B pronouns cooc- 
curring in a domain of binding. 
The strategy of considering the more plausible 
antecedent choices first does not eliminate inter- 
dependency collisions in general, and, moreover, 
940 
1. For each anaphoric NP Y, determine the set of possible antecedents X: 
(a) Verify morphosyntactic or lexical agreement with X (congruence in person, number, and gender, 
lexical recurrence ete, depending on the type of Y) 
(b) If the antecedent candidate X is intru.sentential, check whether the binding l)rinciples of Y and X 
are satisfied: for the proposed eoindexing, 
i. verify that the binding principle of Y is satisfied constructively, 
ii. verify that the binding principle of X is not violated. 
(c) If Y is a type B 1)ronoun, antecedent candidate X is intr~scntential, and, according to surfa(:e 
order, X follows Y (i.e. tit(.' resumption would be cataphorie), verify that X is definite. 
2. Plausibility scoring and sorting: 
(a) For each surviving pair (Y/, A~) of anaphor and antecedent candidate: deterinine the munerical 
plausibility score v(Y/, Xj), which ranks Xj relatively to Y/, based on case role inertia, recency, 
cataphor penalty, and subject prefe.rence, deI)ending on the type of I~. 
(b) (local sorting) For each anai)hor Y: sort their individual antecedent cm, didates Xj according to 
decreasing plausibility v(Y, X a ). 
(c) (global sorting) Sort the anaphors V according to decreasing I)lausibility of their individual best 
antecedent candidate. 
Antecedent Selection: Consider anaphors Y in the order determined in step 2c. Suggest antecedent 
candidates X(Y) in the order determined in step 213. Select X(Y) as candidate if there is no interde- 
pendency, i.e. if 
(a) the morphosyntactic featmes of Y and X(Y) are still compatible% 
(b) for each NP Z whose coindexing with X(Y) has been determined in the current invocation of 
the anaphor resolution algorithm: the coindexing of Y anti Z which results as a side effect when 
chosing X(Y) as antecedent for Y does not violate the binding 1)rincil)les. 
To allow for m, efficient detection of intex(let)endencies , store the selected antecedent separately fl'om 
corefercnt occurrences contributed by earlier invocations of the algorithm. 
Figure \]: The KONTEXT Anaphor Resolution Algorithm 
3. 
does not guarantee that the global maximum of 
plausibility is reached. Because of its practical 
performance, however, it proved to be a satisfac- 
tory substitute for the generate-and-test strategy. 
5 Exploring the Limits 
The determination of the substructure describing 
a local domain iv not always easy. Whereas for 
NPs with possessive markers (of. example (10)) 
the matter tends to be clear, a common source of 
difficulties emerges from adjectivally used partici- 
ples and from deverbative NPs. In the latter case, 
e.g. a genitival attribute may instantiate, depend- 
lug on the NP, either the subject (.qenitiwts sub- 
jectivus) or the object (.qenitiwts objectivus) (for 
German, cf. (Teubert, 1979)). As the following 
examl)les demonstrate, it iv insufficient to know 
merely about the existence of a h)cal domain. In 
general, it is necessary to determine the instantia= 
tion of its participants, but this, at least in certain 
<:ases, involves pragmatic inferencing. 
(lla) Pauli accepts the decision for himi. 
(llb) * Pauli accepts the decision for himselfi. 
According to accet)tability judgements, decision 
introduces a local binding domain. But a change 
6In German, this kind of interdependency may 
arise, due to lnorphosyntactic ambiguity, in case of 
multiple occurrences of the pronoun sic. 
of tile matrix clause verb leads to a different judge- 
ment, while tile syntactic structure is preserved: 
(12a) Pauli revises the decision for himi. 
(12b) Pauli rcviscs the decision for himselfi. 
The clue lies in the observation that a pragmatic 
restriction is governing the instantiation of the 
implicit local subject in exmnples (11), but not 
in examples (12). in (11@, duc to the obvious 
conclusioil that someone who accepts an action is 
not the conscious actor of it, .Paul is pragmati- 
(:ally ruled out as the local subject of the decision 
domain. On the other hand, revise leaves open 
whether Paul or someone else is the decider. This 
explanation is confirmed by the following data: 
(13a) Paull revises Sarn'sj decision for h, imi. 
(13b) * Pauli revises Sam 'sj decision fi)r himselfi . 
(13e) *Pauli revises hisi decision for himi. 
(13d) Pauli revises hisi decision for himselfi. 
Current approaches (Strube and Hahn, 1995; Lap- 
pin amt Leass, 1994) ignore this subtlety by 
merely taking into account NP domains Which are 
established by possessive determiners. As a con- 
sequence, wrong results may be obtained, e.g. in 
case of example (lla), as there is no t)ossessive 
modifier, Paul will not be considered to be an mt- 
tecedent candidate for him. With these difficulties 
in mind, questionable antecedent decisions may t)e 
941 
marked as depending on particular local instanti- 
ations, by this means providing a starting point 
for more comprehensive considerations which take 
into account the relation between structural re- 
strictions and the resolution of ellipsis. 
6 Conclusion 
Starting with a recapitulation of current work on 
anaphor resolution, it was argued for an approach 
which bases on syntactic restrictions. 
The original formulation of Chomsky's Binding 
Theory proved to be unsuitable for immediate im- 
plementation. Straightforward approaches may 
fail in cases in which interdependencies between 
antecedent decisions arise. Based on this observa- 
tion, an algorithm has been presented which, on 
the one hand, is interdependency-sensitive, but, 
on the other hand, avoids computational unfeasi- 
bility l)y following a strategy according to which 
the choices with the highest plausibility are con- 
sidered first. For each decision, its dynamic com- 
patibility with the earlier (more plausible) deci-. 
sions is verified. The practical behaviour of the 
algorithm fulfilled the expectations. 
There are, however, limitations to the scope of 
syntactic constraints. It has been demonstrated 
that, in general, the construction of appropriate 
representations for binding domains may necessi- 
tate semantic or pragmatic inferencing. 
A topic which should be subject of further re- 
search is the interdependency between parse tree 
construction and anaphor resolution. Up to now, 
it has been assumed tacitly that, at the time of 
binding constraint application, the surface struc- 
ture representation is available. The construction 
of this representation involves disambiguation de- 
cisions (relative clause attachment, prepositional 
phrase attachment, and uncertainty of syntactic 
flmction), which, due to their structure determin- 
ing effects, may interfere with the antecedent op- 
tions of anaphor resolution (cf. (Stuckardt, 1996)). 
At current, the KONTEXT text analysis system 
employs a processing model according to which 
parsing is performed prior to anaphor resolution. 
Because of the interdependency between parsing 
and anaphor resolution, however, these two prob- 
lem (:lasses should be handled at one stage of pro- 
cessing rather than sequentially. 
References 
Jaimc G. Carbonell, Ralf D. Brown. 1988. Ana- 
phora Resolution: A Multi-Strategy Approach. 
In: Proc. I2th Int. Conf. on Computational 
Linguistics, COLING, Vol. I, 96-101, Budapest. 
David Carter. 1987. Interpreting Anaphors in 
Natural Language Texts. Ellis Horwood Series 
in Artificial Intelligence, Chichester. 
Noam Chomsky. 1981. Lectures on Government 
and Binding. Foris Publications, Dordrecht. 
Noam Choinsky. 1986. Knowledge of Language. 
Praeger, New York. 
Nelson Correa. 1988. A Binding Rule \]or Go- 
vernment-binding Parsing. In: Proc. 12th Int. 
Conf. on Computational Linguistics, COLING, 
Vol. I, 123-129, Budapest. 
Barbara J. Grosz, Candace L. Sidner. 1986. At- 
tention, intention and the Structure of Dis- 
course, in: Computational Linguistics, Vol. 12, 
Number 3, 175-204. 
Karin Haenelt. 1994. Das Textanalysesystem 
KONTEXT. Konzcption und AnwendungsmSg~ 
lichkciten. In: Sprache und Datcnvcrarbeitung, 
18, 1, 17-31, 1994. 
Jerry Hobbs. 1978. Resolving Pronoun Refer- 
enccs. In: Lingua, 44,311-338. 
Susmnu Kuno. 1987. Anapho~t and Discourse 
Principles. In: Makoto Nagao (ed). Language 
and Artificial Intelligence. Elsevier, 1987. 
Vittorio M. Lampugnani (ed). 1983. Lexikon 
der Architektur des 20. ,Iahrhundcrts. Hatje, 
Stuttgart. 
Shalom Lappin and Herbert J. Lea~s. 1994. An 
Algorithm for P~vnominal Anaphora Resolu- 
tion. In: Computational Linguistics, 20 (4), 
535-561. 
Candace L. Sidner. 1983 I,bcusing in the Com- 
prehension of Definite Anaphora. In: Michael 
Brady, Robert C. Berwick (eds). Computational 
Models of Discourse. M.I.T. Press, Cambridge, 
MA, 267-330. 
Michael Strube, Udo Hahn. 1995. ParscTalk 
about Sentence- and Text-Level Anaphora. In: 
Proc. ~th Conf. of the European Chapter of 
the Association for Computational Linguistics, 
EACL, 237-244. Dublin, March 28-31, 1995. 
Roland Stucl~trdt. 1996. An Interdependency- 
Sensitive Approach to Anaphor Resolution. To 
appear in: Proc. of DAARCg6 - Discourse 
Anaphora and Anaphor Resolu, tion Colloquium,. 
Lancaster University, 17-18th July, 1996. 
Wolfgang Teubert. 1979. Valenz des Sub- 
stantivs. Attributive Ergiinzungen nnd Angaben. 
P~idagogischer Verlag Schwann. 
Bonnie L. Webber. 1989. Deictic Reference and 
Discourse Structure. Department of Computer 
and Information Science, University of Pennsyl- 
vania, Philadelphia PA 19104-6389. 
942 
