An Algorithm to Co-Ordinate Anaphora Resolution and PPS 
Disambiguation Process 
Saliha Azzam 
CRIL Ing6nierie - CAMS La Sorbonne 
174, Rue de la RSpubhque 
92817 Puteaux 
France 
azzam@cril-ing.fr 
Abstract 
Both anaphora resolution and preposi- 
tional phrase (PP) attachment are the 
most frequent ambiguities in natural lan- 
guage processing. Several methods have 
been proposed to deal with each phe- 
nomenon separately, however none of 
proposed systems has considered the way 
of dealing both phenomena We tackle 
this issue here, proposing an algorithm 
to co-ordinate the treatment of these two 
problems efficiently, i.e., the aim is also 
to exploit at each step all the results that 
each component can provide. 
1 Introduction 
Several methods have been proposed to deal 
with anaphora resolution and prepositional phrase 
(PP) attachment phenomenon and separately, so 
that the literature is very abundant : for PPs see 
e.g., (Frazier and Fodor, 1979; Hobbs, 1990), and 
(Wilks and Huang, 1985), and for anaphora see 
e.g., (Carter, 1986; Reinhart, 1983) and (Sidner, 
1983). However none of these methods has con- 
sidered the way of dealing both phenomena in the 
same concrete system. 
We propose in this paper an algorithm that 
deals with both phenomena, in the same anal- 
yser. The anaphora module pertains to the recent 
methods, uses a set of resolution rules based on 
the focusing approach, see (Sidner, 1983). These 
rules are applied to the conceptual representation 
and their output is a set of candidate antecedents. 
Concerning the PPs, unattached prepositions in- 
volve empty or unfilled roles in the Conceptual 
Structures (CSs), expressed in a frame-based lan- 
guage (Zarri, 1999). The disambiguation proce- 
dure alms at filling the empty roles using attach- 
ment rules. 
This work was accomplished in the context of 
COBALT project (LRE 61-01 ), dealing with fi- 
nancial news. A detailed discussion about both 
procedures of anaphora resolution and PP attach- 
ment is largely developed in (Azzam, 1994). 
2 The algorithm 
Two of the main principles of the algorithm are : 
a) The algorithm is applied on the text sentence 
by sentence, i.e. the ambiguities of the previous 
sentences have already been considered (resolved 
or not). 
b) The anaphora procedure skips the resolution 
of a given anaphor when this anaphor is preceded 
by an unattached preposition. This is because the 
resolution rules may have an empty role as a pa- 
rameter, due to this unattached preposition. The 
resolution of the anaphor is then postponed to the 
second phase of anaphora resolution. 
The proposed procedure is based on successive 
calls to the anaphora module and to the PP at- 
tachment module. The output of each call is 
a set of CSs that represent the intermediate re- 
suits exchanged between each call and on which 
both modules operate in turn. The aim is to fill 
the unfilled roles in the CSs, due to anaphora or 
unattached PPs. To summarize the algorithm is: 
1) Apply the anaphora module first. 
2) Apply the PP attachment procedure. 
3) If some anaphora.are left unresolved, apply 
the anaphora module again. 
4) If there are still unattached PPs, apply the 
attachment procedure again. 
5) Repeat (3) and (4), until all VPs and 
anaphors are treated. 
The order in which the two modules are called is 
based on efficiency deduced from statistical data 
performed on COBALT corpuses. 
Three main cases are faced by the algorithm : 
a) When the anaphor occurs before a given prepo- 
sition in the sentence, its resolution does not de- 
pend on where the preposition is to be attached 
(except for cataphors that are quite rare). In this 
284 
case the anaphora module can be applied before 
the attachment procedure. 
The example 1 below shows that the resolution 
of the anaphoric pronoun that must be performed 
first and that the PP starting with of be attached 
later. 
(1) The sale of Credito was first proposed last 
August and that of BCI late last ~lear. 
b) When the anaphor occurs after one or sev- 
eral unattached prepositions, it could be an intra- 
sententiai anaphor (i.e. referring to an entity in 
the same sentence), then its resolution may de- 
pend on one of the previous prepositional phrases. 
In this case, the resolution of the anaphora is post- 
poned to a next call of the anaphora module ac- 
cording to principle b) stated above. 
c) When the anaphor is included in a PP (particu- 
lar case of b), PP attachment rules need semantic 
information about the "object" of the PP; when it 
is a pronoun, no semantic information is available, 
so that the attachment rules can not be applied. 
The anaphoric pronouns have to be resolved first, 
so as to determine what semantic class they re- 
fer to ; the PP attachment procedure can then 
be applied. When a sequence contains more than 
two such PPs, i.e., with anaphors as objects, the 
length of a cycle is more than 4. 
3 An example 
(~) UPHB shares have been suspended since Oc- 
tober ~g at the firm's request following a surge in 
its share price on a takeover rumour. 
- The pronoun its can not be resolved by the 
anaphora resolution module because it is preceded 
by unattached PPs ; its resolution is skipped. 
- The PP attachment procedure is then called to 
determine the attachment of since and at while 
the object of the in PP comprises an anaphoric 
pronoun its (case c). 
- The anaphora module is called again to resolve 
the anaphoric pronoun its, which is possible, in 
this example, since the previous PPs have been 
attached and there is no anaphors before. 
- Finally, the PP attachment procedure has to be 
called again for the in PP. 
Notice that even if each module is called sev- 
eral times, there is no redundancy in the process- 
ing. The algorithm should be considered as the 
splitting of both anaphora resolution and PP at- 
tachment procedures into several phases and not 
as the repetition of each procedure. 
4 Conclusion 
The objective was to emphasise more than it has 
been done until now, the fact that PP attach- 
ment and anaphora resolution could interact in 
the same system in order to produce a complete 
conceptual analysis, instead of slowing down each 
other. The algorithm we proposed in this. paper, 
is independent of the used approaches in both 
anaphora and attachment modules. It concerns 
rather the way of managing the interaction be- 
tween the two modules. 
Our actual work addresses more the problems 
inside each module. The attachment module has 
been implemented at 99%. Presently we are work- 
ing on the extension of the anaphora module par- 
ticularly to deal also with the anaphoric definite 
noun phrases. 
References 
Azzam, S. 1994. CLAM COBALT conceptual 
analyser (COBALT Tech. Report Del6.2). CRIL 
Ing~nierie. 
Carter, D. 1987. Interpreting Anaphors in nat- 
ural language Texts. Chichester: Ellis Horwood. 
Frasier, L. and Fodor, J. 1979. The sausage 
machine: A New Two-Stage Parsing Model, Cog- 
nition, 6. 
I-Iobbs, J.R., and Bear, J. 1990. Two Princi- 
ples of Parse Reference in Proceedings of the 13th 
International Conference on Computational Lin- 
guistics - COLING/90, vol. 3, Karlgren, H., ed. 
Helsinki: University Press. 
Reinhart, T. 1983. Anaphora and Semantic In- 
terpretation. London : Croom Helm. 
Sidner, C.L. 1983. Focusing for Interpretation 
of pronouns. American Journal of Computational 
Linguistics, 7, 217-231. 
Wilks, Y., Huang, X., and Fass, D. 1985. Syn- 
tax, Preference and Right Attachment, IJCAI. 
Zarri, G.P. 1992. The descriptive compo- 
nent of hybrid knowledge representation language, 
In: Semantic networks in Artificial Intelligence, 
Lehmann, F., ed. Oxford: Pergamon Press. 
285 
