Resolving Bridging References in Unrestricted Text 
Massimo Poesio, Renata Vieira and Simone Teufel 
Centre for Cognitive Science - University of Edinburgh 
2, Buccleuch Place EH8 9LW Edinburgh UK 
{poesio, renata, simone}@cogsci, ed. ac. uk 
Abstract 
Our goal is to develop a system capable 
of treating the largest possible subset of 
definite descriptions in unrestricted writ- 
ten texts. A previous prototype resolved 
anaphoric uses of definite descriptions and 
identified some types of first-mention uses, 
achieving a recall of 56%. In this paper 
we present the latest version of our system, 
which handles some types of bridging refer- 
ences, uses WordNet as a source of lexical 
knowledge, and achieves a recall of 65%. 
1 Previous Work 
We are in the process of developing a system for 
interpreting definite descriptions (DDs) in written 
text without restrictions of domain. The implemen- 
tation work has been supported by an analysis of 
definite description use in corpora of written lan- 
guage (Poesio and Vieira, 1997). In one of our ex- 
periments, we asked 2 subjects to classify the uses of 
definite descriptions in a corpus of English texts 1 us- 
ing a taxonomy derived from the proposals of (Clark, 
1977; Hawkins, 1978; Prince, 1981; Fraurud, 1990; 
Prince, 1992). In the taxonomy used in that study, 
we defined bridging references as those uses of defi- 
nite descriptions based on previous discourse which 
require some reasoning in the identification of their 
textual antecedent (rather than just matching iden- 
tical nouns). These definite descriptions may be co- 
referential with an entity already introduced in the 
discourse, but be characterized by a different head 
noun (as in a car.., the vehicle); or may be simply 
semantically related to it (in the sense that the door 
is related to house). Of the 1040 DDs in that cor- 
pus, 204 (20%) were identified as bridging descrip- 
tions, 312 (30%) as anaphoric (DDs and antecedents 
1A set of randomly selected parsed articles fl'om the 
Wall Street Journal contained in the ACL/DCI CD-ROM. 
which co-refer and have the same head noun), and 
492 (47%) as larger situation/unfamiliar (Hawkins, 
1978) (Prince's discourse new (Prince, 1992)); the 
remaining definite descriptions were classified as id- 
iomatic or doubtful cases. 
These results led us to concentrate initially on re- 
solving same-head anaphoric DDs and on recognis- 
ing larger situation/unfamiliar uses. Our analysis of 
the corpus suggested that many of the latter could 
be recognised using syntactic heuristics: e.g., on the 
basis of the presence of restrictive pre- and post- 
modification, of the presence of special predicates 
(such as the superlatives first, best), or because the 
DD occurred in a copula or appositive construction. 
A first prototype with these capabilities (Vieira 
and Poesio, 1997) achieved an overall recall of 56% 
and precision of 84% when tested on our corpus. Of 
all anaphoric DDs 72% were resolved, and 74% of all 
larger situation and unfamiliar uses were identified. 
The definite descriptions not handled by this first 
prototype were typically larger situation uses based 
on common knowledge (such as the government) and 
bridging descriptions. In this paper we present our 
subsequent work devoted to handling some of these 
remaining cases. 
2 Bridging Descriptions 
Linguistic and computational theories of bridging 
descriptions identify two main subtasks involved in 
their resolution: first, finding the element in the 
text to which the bridging description is related 
(ANCHOR) and second, finding the relation (LINK) 
holding between the bridging description and its an- 
chor (Clark, 1977; Sidner, 1979; Helm, 1982; Carter, 
1987; Fraurud, 1990; Strand, 1997). 
A speaker is licensed to use a bridging DD when 
he/she can assume that the common-sense knowl- 
edge required to identify the relation is shared by 
the listener (Hawkins, 1978; Clark and Marshall, 
1981; Prince, 1981). This reliance on commonsense 
knowledge means that, in general, a system could 
only resolve bridging references when supplied with 
an adequate knowledge base; for this reason, the 
typical way of implementing a system for resolving 
bridging references has been to restrict the domain 
and feeding the system with hand-tailored world 
knowledge. (This approach is discussed in detail in 
(Carter, 1987)). In order to get a system capable of 
performing on unrestricted text, we decided to use 
WordNet (WN) (Miller, 1993) as an approximation 
of a knowledge base containing generic information, 
and to supplement it with heuristics to handle those 
cases which WN couldn't handle. 
Vieira and Teufel (1997) analyzed the corpus to 
identify the cases of bridging descriptions that could 
be resolved using WordNet, those for which we could 
use heuristics, and those that couldn't be interpreted 
at the moment. Six classes of bridging descriptions 
were identified. 2 
Synonymy/Hyponymy/Meronymy This class 
(henceforth, Syn/Hyp/Mer) includes those DDs 
which are in a synonymy/hyponymy/meronymy re- 
lation with their anchors, i.e., the kind of seman- 
tic relation that is currently encoded in WordNet. 
Examples 3 are: 
(1) a. Synonymy: new album -- the record; three 
bills -- the legislation. 
b. Hypernymy-Hyponymy: rice -- the plant; 
the daily television show -- the program. 
c. Meronymy (part-of relation): plants -- the 
pollen; house -- the chimney. 
Names This class includes definite descriptions 
that refer back to proper names such as people's and 
company names, as in: 
(2) Bach -- the composer; Pinkerton's Inc -- the 
company. 
Compound Nouns This class includes bridging 
descriptions whose LINGUISTIC ANCHOa (i.e., the el- 
ement in the text to which they are related) is a 
noun occurring as part of a compound noun other 
than the head. Examples include: 
(3) stock market crash -- the markets; discount 
packages -- the discounts. 
Events These are cases where the linguistic an- 
chors of DDs are not NPs but VPs or sentences. 
Examples are: 
2We should stress that this classification is primarily 
motivated by processing considerations. 
3Note that the examples in (la) and (lb) are classified 
as bridging even though the relation is of co-reference. 
Class Total 
Syn/Hyp/Mer 12/14/12 
Names 49 
25 Compound Nouns 
Events 40 
Discourse Topic 15 
Inference 37 
Total 204 
% 
19% 
24% 
12% 
20% 
7% 
18% 
100% 
Table 1: Distribution of types of bridging DDs 
(4) Individual investors and professional money 
managers contend. -- They make the argu- 
ment ...; Kadane Oil Co. is currently drilling 
two wells and putting money into three others. 
-- The activity ... 
Discourse Topic There are some cases of DDs 
which are related to the (often implicit) discourse 
topic (in the sense of (Reinhart, 1981)) of a text, 
rather than to some specific NP or VP. For instance, 
(5) the industry (in a text whose discourse topic 
is oil companies); the first half(in a text whose 
discourse topic is a concert). 
Inference We collect in this class all the cases of 
bridging descriptions whose relation with their NP 
anchor was based on more complex inferential rela- 
tions: for example, cases in which the relation be- 
tween the anchor and the DD was of reason, cause, 
consequence, or set-membership: 
(6) last week's earthquake -- the suffering people 
are going through; Democratics/Republicans 
-- the two sides. 
The relative importance of these classes in our cor- 
pus is shown in Table 1. This classification is based 
on what we took to be the main linking relation for 
each of the 204 bridging DDs in the corpus 4. 
3 Resolution of Bridging 
Descriptions 
We used Vieira and Teufel's analysis as the basis for 
the implementation of a second prototype. In this 
section we discuss how this prototype handles the 
different types of bridging descriptions. 
3.1 Bridging DDs and WordNet 
We implemented a WordNet interface that reports a 
possible semantic link between two nouns when one 
4One problem with bridging references is that they 
are often related to more than one antecedent in the 
discourse (Poesio and Vieira, 1997). 
of the following is true: 
* The nouns are in the same synset (= synonyms 
of each other), as in suit -- lawsuit. 
• The nouns are in direct hyponymy relation with 
each other, for instance, dollar -- currency. 
• There is a direct or indirect meronymy between 
them. Indirect meronymy holds when a con- 
cept inherits parts from its hypernyms, like 
car inherits the part wheel from its hypernym 
wheeled_vehicle. 
• Due to WordNet's idiosyncratic encoding, it is 
often necessary to look for a semantic relation 
between coordinate sisters, i.e. hyponyms of the 
same hypernym, such as home -- house which 
are hyponyms of housing, lodging. 
Sometimes, ifa relation between two head nouns is 
not encoded in WN directly, the semantic closeness 
might be found through the compound nouns made 
up of them. Thus, for a pair such as record, album 
we find synonymy between record_album and album. 
We ran a test in which WordNet was used to iden- 
tify the DD's anchors. For each of the 204 bridging 
DDs in our corpus, we considered the NPs in the 
previous five sentences as a potential anchor 5, and 
queried WN with the DD and potential anchor pair. 
WordNet reported a possible relation for 107 of 
the 204 DDs. Often, more than one link was found 
between a DD and a potential anchor, 6 and some- 
times links could be found with more than one po- 
tential anchor in the previous five sentences. Only 
in 34 of these 107 cases we found at least one appro- 
priate anchor linking relation, and only in 21 cases 
we found only appropriate anchors (for 13 there was 
a mixture of suitable and unsuitable anchors). 
Of these 34 DDs for which a correct anchor was 
found, only 18 were among those we had classified 
as Syn/Hyp/Mer. In 8 of these 18 cases, WordNet 
found a link with an anchor that was not the one we 
had identified manually, but which was still valid; 
for instance, we identified the link pound -- the cur- 
rency which was not found in WordNet, whereas our 
automatic search found sterling -- the currency). 
5Our system does not currently include a proper seg- 
mentation algorithm. Instead, we use a simple recency 
heuristic--we only consider the antecedents in the n pre- 
vious sentences, where n is a constant determined empir- 
ically. In our previous work (Vieira and Poesio, 1997). 
we observed that 5 was the value of n which gave the 
best tradeoff between precision and recall. 
6For instance, transaction and trade were reported to 
be in a hypernym relation and were also reported as coor- 
dinate sisters having as common hypernymy commerce. 
The 16 remaining relations were found for DDs that 
we had not classified as Syn/Hyp/Mer: for instance, 
whereas we had classified the DD the population as 
belonging to the class of DDs based on names hav- 
ing as anchor Koreans, our implementation found a 
WordNet relation for the pair nation -- the popu- 
lation; the system also found a few relations with 
proper names, such as Bach -- the composer. 
In the following tests we have considered only di- 
rect meronymy, as indirect meronymy presented ex- 
treme low recall and precision at a very expensive 
computational cost. 
In order to reduce the number of false positives 
(86 out of 107) 7, we tried using a stack-based ap- 
proach towards finding potential anchors in the pre- 
vious sentences, as suggested in (Sidner, 1979); i.e., 
the system would go back one sentence at a time, 
and stop as soon as a relation with a potential an- 
chor was found, rather than trying to find all possible 
links. As a result, we found exactly one correct an- 
chor for 30 DDs, slightly improving our results (76 
false positives). 
Class Total Right Wrong P 
Syn 11 4 7 36% 
Hyp 59 18 42 30% 
Mer 6 2 4 33% 
Sister 30 6 24 20% 
Total 106 30 76 28% 
Table 2: Analysis of the anchors found in WN 
Table 2 shows the distribution of the different se- 
mantic relations between DDs and the anchors found 
by our stack-based search. It presents precision fig- 
ures (P) related to each type of relation s. Sister 
relations are the least satisfactory ones. 
We tested, in particular, whether WordNet en- 
coded a semantic link between the 38 syn/hyp/mer 
relations in our corpus (just described) plus other 32 
relations extracted from a second corpus study (Poe- 
sio and Vieira, 1997). The results for the 70 DDs are 
summarised in Table 3. Overall recall (R) was 46%. 
We could have expected 100% precision, since we 
had manually identified the anchors, but the preci- 
sion figures (P) report an error when a sister relation 
is found instead of the expected (syn/hyp/mer) re- 
7The anchors found for 73 of the DDs were incor- 
rect, for the remaining 13 DDs the resulting anchors were 
mixed (some right/some wrong). 
SWe cannot estimate recall since we do not have a 
precise number of syn/hyp/mer anchors that should be 
found (as different types of anchors may allow resolution 
of bridging descriptions). 
lation. The poorest recall was surprisingly obtained 
for synonymy relations, followed by meronymy re- 
lations, as expected, since these are only partially 
implemented in WordNet. 
\[ Class I Total In WN Out R P 
Syn 20 7 13 35% 71% 
Hyp 32 18 14 56% 94% 
Mer 18 7 11 38% 71% 
Total 70 32 38 46% 84% 
Table 3: Search for semantic relations in WN 
The low recall for synonymy relations may be 
due to the context dependent, specialized senses of 
sublanguage terminology (for instance, crash, bust 
and slump in Economics terminology). Some nouns 
were not even encoded in WN (such as newsweekly, 
spino3~). Other relations were missed due to the 
unexpected way in which knowledge is organised in 
WordNet. For example, no association was found be- 
tween house and walls, because house is not encoded 
in WordNet as a hyponym of building but of housing, 
and housing does not have a meronymy link to wall 
whereas building does. Another example of counter- 
intuitive position in the hierarchy is that of catastro- 
phe, not listed as a hypernym of earthquake, but as 
its coordinate sister. This example demonstrates the 
problems that WordNet lexicographers faced when 
they had to coerce real-world concepts into a hierar- 
chy with single, no-default inheritance: catastrophes 
clearly subclassify into earthquakes and floods, but 
minor earthquakes don't have to be catastrophic. 
3.2 Proper names 
Definite descriptions which refer back to proper 
names are very common in Wall Street Journal arti- 
cles. Processing such DDs involves, first, determin- 
ing an entity type for each name in the text, then 
searching for semantic relations. If we get the entity 
type person for the a name such as Mrs. Y.J. Park 
we could, ideally, resolve the subsequent DD the 
housewife using WordNet. A few names are avail- 
able in WordNet: typically, famous people, coun- 
tries, states, cities and languages. Other entity types 
can be identified using appositive constructions and 
abbreviations like Mr., Co., Inc. etc. as cues. 
The algorithm we developed, based on a mixture 
of access to WordNet and heuristics such as those 
we described, found the correct type for 66% of the 
names in our corpus (535/814). Including a back- 
tracking mechanism which re-processes a text look- 
ing for missing name types (with this mechanism we 
identify the type for the name Morishita in a textual 
sequence like Morishita -- Mr. Morishita) increases 
our recall to 69% (562/814). We then used WordNet 
to match the types found with previous references 
in the text. This resulted in the resolution of 53% 
(26/49) of the cases based on names. We missed rela- 
tions which are not found in WordNet (for instance, 
Mr. Morishita -- the 57 year-old). But again we 
also found a large number of false positives. 
3.3 Compound Nouns 
Sometimes, a bridging description is linked to a non- 
head noun in a compound noun: 
(7) stock market crash -- the markets; rule 
changes -- the rules; discount packages ~ the 
discounts. 
One way of processing these definite descriptions 
would be to update the discourse model with dis- 
course referents not only for the NP as a whole, 
but also for the embedded nouns: for example, after 
processing stock market crash, we could introduce a 
discourse referent for stock market, and another dis- 
course referent for stock market crash. The DD the 
markets would be co-referring with the first of these 
referents (with identical head noun), and we could 
simply use our anaphora resolution algorithms. This 
solution, however, makes available discourse refer- 
ents that are generally unaccessible for pronominal 
anaphora. We therefore followed a different route: 
our algorithm for identifying antecedents attempts 
to match not only heads with heads, but also the 
head of a DD with the pre-modifiers of a previous 
NP, the pre-modifiers of a DD with the pre-modifiers 
of its antecedents, and the pre-modifiers of the DD 
with the head of a previous NP. With this, we ob- 
tained a total of 54 resolutions of which 36 were cor- 
rect, although we did not always find a main linking 
relation licensing the use of a DD. 
Examples of correct resolutions are: 
(8) a. Head of DD with pre-modifier of an- 
tecedent: the stock market crash -- the 
markets; rule changes -- the rules; 
b. Pre-modifiers of DD with pre-modifiers of 
antecedent: most oil companies -- the oil 
fields; his art business -- the art gallery; 
c. Pre-modifiers of DD with head of an- 
tecedent: New ~brk City -- the city coun- 
cil district lines; a 15-acre plot and main 
home - the home site. 
There are also cases in which the pre-modifiers 
plus the head noun of a DD may indicate a bridg- 
ing reference: we may find an antecedent with the 
same head noun for a DD but referring to a different 
entity, this being signalled by the pre-modification. 
Some examples: 
(9) the company's abrasive segment -- the engi- 
neering materials segment; Italy's unemploy- 
ment rate -- the southern unemployment rate; 
Pinkerton -- the new Pinkerton; increases of 
3.9 ~ -- the actual wage increases may have 
been bigger. 
Our previous heuristics for treatment of pre- 
modifiers in anaphoric resolution handled the first 
two examples correctly (Vieira and Poesio, 1997): 
as they present different pre-modifiers we did not 
treat them as anaphoric in the first version of our 
system. Such cases, as well as DDs modified by new 
and actual (last two examples), may now be treated 
as bridging references 9. 
3.4 Events 
To process DDs based on events (situations or propo- 
sitions), we are trying, as a first approach, to trans- 
form verbs into their nominalizations, and then look- 
ing for a relation in WordNet. Some nominalizations 
can be generated by general procedures or learned 
by means of a stochastic method: e.g., we could use 
WordNet's morphology component as a stemmer, 
and augment the verbal stems with the most com- 
mon suffixes for nominalizations which could be kept 
in a list, like -ment, -ion. In our corpus, 17% (7/40) 
of the bridging references based on events are direct 
nominalizations of this type (for instance, changes 
were proposed - the proposals). 
Another 17% are cases in which knowledge of the 
semantics of the verb is necessary (as in borrowed - 
the loan). The remaining 66% (26 cases) of bridg- 
ing DDs based on events require inference reasoning 
based on the compositional meaning of the phrases 
(as in It went looking for a partner - pitching the 
prospect); these cases are out of reach just now, as 
well as the cases listed under "discourse topic" and 
"inference". 
4 Restrictive Post-modification as 
Anchors 
Whereas the problem of finding the appropriate tex- 
tual anchor for bridging descriptions requires knowl- 
edge inference and reasoning, DDs with restrictive 
post-modification give the reader both anchor and 
description in the same expression. As Hawkins 
points out (Hawkins, 1978): 
... first mention uses of the with both ref- 
erent establishing relative clauses and as- 
9This idea is not implemented yet. 
sociative clauses are not essentially differ- 
ent from the other uses mentioned in the 
last section (Hawkin's associative uses/our 
bridging uses). The only difference is that. 
in the latter uses set identification and tile 
locatability of the referent were possible 
on account of previous triggers, whereas 
it is now function of the modifier itself 
to provide the information which makes 
set identification and location possible- 
(parentheses ours). 
This could be seen as an advantage, as we directly 
find the anchors of these DDs. They are quite com- 
mon uses of DDs, but not much attention has been 
devoted to them as a special case of anchor linking 
relations. The main problem for these cases is to find 
out their links, which is also a remaining problem for 
our proposals of anchor identification throughout t he 
paper. 
The head noun of a DD and its modifier may be 
related in different ways, as shown by the examples: 
(10) the number of job seekers; the anthers of the 
plant; the ideal of a level playing field; the flip 
side of the Stoltzman personality. 
In some cases, the relation between noun and com- 
plement seem to be looser than the relations for 
bridging descriptions. Sequences such as the laws of 
heredity; the cost of the plan are acceptable, whereas 
heredity -- the laws; the plan -- the cost are unlikely 
to occur. On the other hand, bridging such as the 
house -- the kitchen; the firm -- the owners are as 
acceptable as the kitchen of the house; the owners of 
the firm. Some proposals of a systematic treatment 
for the identification of anchor linking relations for 
bridging DDs are (Heim, 1982; Barker, 1991; Poesio, 
1994; Strand, 1997). 
5 Discussion 
We presented our most recent results concerning the 
resolution of bridging descriptions. We identified 
different types of bridging descriptions, and we pro- 
posed a treatment for each of them separately. 
We presented some preliminary experiments with 
WordNet. Our experience was mixed. WordNet was 
useful in determining the type of entity for some 
of the proper names in our corpus, typically cities, 
states and countries. On the other hand, WordNet 
proved to be unreliable for the automatic identifi- 
cation of possible link relations, as too many false 
positives were found. A 'blind' WordNet search for 
semantic relations is also very expensive computa- 
tionally. A mechanism for focus tracking (Grosz and 
Sidner, 1986) or a clustering algorithm should be ap- 
plied first in order to minimise the costs. 
In order to have proper names available for reso- 
lution of future references, it is useful to create dis- 
course referents for them which contain their entity 
types. Up to now we have identified an entity type 
for 69% of the names in our corpus, and we resolved 
53% of the DDs referring back to proper names with 
the help of WordNet. 
Pre-modifiers in compound nouns may license the 
use of definite descriptions. We have presented some 
preliminary tests that should be further developed. 
Cases of bridging references based on events or 
propositions usually involve common-sense reason- 
ing; some of them (in our corpus, 34% of all cases 
based on events) can however be solved by trans- 
forming verbs into their nominalizations and then 
searching for a semantic relation. 
We also claimed that the same problem of deter- 
mining a linking relation for bridging descriptions 
holds for first mention uses of DD based on restric- 
tive post-modification. 
As an estimate, we could say that about 60% of 
the cases in the bridging class could be treated by 
developing the ideas proposed here. We combined 
the proposed heuristics with the first version of our 
system-- we applied the heuristics for proper names, 
compound nouns and WordNet consult (in this or- 
der) to the cases which the system did not handle 
before. The impact on the overall system perfor- 
mance was an increase in recall from 56% to 65% 
(note that the bridging class is a small class com- 
pared to the others) but precision decreased from 
84% to 82%. The heuristics should be further devel- 
oped, and their integration into the system should be 
worked out in more detail--the heuristics could be 
implemented in parallel or through a decision tree. 
Acknowledgements The authors would fike to 
thank Kjetil Strand and the anonymous referees for their 
comments on cartier drafts of the paper. The authors are 
supported by an EPSRC Advanced Fellowship, a CNPq 
studentship and an EPSRC studentship, respectively. 

References 
Barker, C. 1991. Possessive Descriptions. Ph.D. 
thesis, University of California at Santa Cruz, 
Santa Cruz, CA. 
Carter, D. M. 1987. Interpreting Anaphors in Nat- 
ural Language Texts. Ellis Horwood. 
Clark, H. H. 1977. Bridging. In P. N. Johnson- 
Laird and P.C. Wason, eds., Thinking: Readings 
in Cognitive Science. Cambridge University Press. 
Clark, H. H. and C. R. Marshall. 1981. Definite 
reference and mutual knowledge. In Elements of 
Discourse Understanding. Cambridge University 
Press. 
Fraurud, K. 1990. Definiteness and the processing 
of NPs in natural discourse. Journal of Semantics, 
7:395-433. 
Grosz, B. J. and C. L. Sidner. 1986. Attention, 
intention, and the structure of discourse. Compu- 
tational Linguistics, 12(3) :175-204. 
Hawkins, J. A. 1978. Definiteness and Indefinite- 
ness. Croom Helm. 
Heim, I. 1982. The Semantics of Definite and In- 
definite Noun Phrases. Ph.D. thesis, University of 
Massachusetts at Amherst. 
Miller, G. et al. 1993. Five papers on WordNet. 
Technical Report CSL Report 43, Cognitive Sci- 
ence Laboratory, Princeton University. 
Poesio, M. 1994. Weak definites. In Proceedings of 
the Fourth Conference on Semantics and Linguis- 
tic Theory, SALT-4. Cornell University Press. 
Poesio, M. and Vieira, R. 1997. A Corpus 
based investigation of definite description use. 
Manuscript, Centre for Cognitive Science, Univer- 
sity of Edinburgh. 
Prince, E. F. 1981. Toward a taxonomy of given- 
new information. In P. Cole, ed., Radical Prag- 
matics. Academic Press, pages 223-256. 
Prince, E. F. 1992. The ZPG letter: subjects, defi- 
niteness, and information status. In S. Thomp- 
son and W. Mann, eds., Discourse description: 
diverse analyses of a fund-raising text. John Ben- 
jamins, pages 295-325. 
Reinhart, T. 1981. Pragmatics and Linguistics: An 
Analysis of Sentence Topics Philosophica, 27(1). 
Sidner, C.'L. 1979. Towards a computational the- 
ory of definite anaphora comprehension in English 
discourse. Ph.D. thesis, MIT. 
Strand, K. 1997. A Taxonomy of Linking Relations. 
Journal of Semantics, forthcoming. 
Vieira, R. and M. Poesio. 1997. Processing def- 
inite descriptions in corpora. In S. Botley and 
M. McEnery, eds., Corpus-based and Computa- 
tional Approaches to Discourse Anaphora. UCL 
Press. 
Vieira, R. and S. Teufel. 1997. Towards Resolution 
of Bridging Descriptions. In Proceedings of the 
A CL Student Session. 
