HIERARCHICAL LEXICAL STRUCTURE AND INTERPRETIVE 
MAPPING IN MACHINE TRANSLATION 
Teruko Mitamura Eric H. Nyberg HI 
Center for Machine Translation Center for Machine Translation 
Carnegie Mellon University Carnegie Mellon University 
Pittsburgh, PA 15213 USA Pittsburgh, PA 15213 USA 
teruko@cs.cmu.edu ehn@cs.cmu.edu 
Abstract 
Large-scale knowledge-based machine translation 
requires significant amounts of lexical knowledge 
in order to map syntactic structures to conceptual 
structures. Tfiis paper presents a framework in 
which lexical knowledge is separated into differ- 
ent levels of representation, which are arranged in 
a hierarchical model based on principles of knowl- 
edge representation and lexical semantics. The pro- 
posed methodology is language-independent, and 
has been used to organize lexical knowledge for 
both English and Japanese. 
1 Introduction 
The basic premise of knowledge-based machine translation is 
that accurate, high-quality translation requires a complete se- 
mantic interpretation of the input text (Carbonell and Tomita, 
1987). Therefore, the analysis and generation components of 
a knowlodge-based MT system must have at least the follow- 
ing functional parts: a grammar for the language, a lexicon 
for the language, a shared set of domain concepts, and rules 
that map syntactic structures onto semantic structures (or vice- 
versa for generation). 
.The goal of our work has been to develop a methodology for 
the hierarchical organization of lexical knowledge (lexical en- 
tries and mapping rules) for knowledge-based MT (Goodman 
and Nirenburg, 1991; Mitamura, 1989). Interpretive Map- 
ping refers to the relationship between predicate conceptual 
structures and syntactic structures, and involves two kinds of 
processes: one is a mapping between grammatical fnnctions 
(e.g., subject, object) and semantic roles (e.g., agent, theme); 
the other is a mapping between words (e.g., naguru 'hit') and 
domain concepts (e.g., *HIT). 
We have developed a shared hierarchical structure for lex- 
ical knowledge which can capture significant linguistic gen- 
eralizations, eliminate rexlundancy, and facilitate both knowl- 
edge acquisition and efficient processing. We have imple- 
mented our hierarchy using FranaeKit, an AI knowledge rep- 
resentation language that supports frames and multiple inher- 
fiance (Nyberg, 1988). 
Our system demonstrates rite integration of a linguistic for- 
realism with a frame-based knowledge representation system. 
We have analyzed a large corpus of Japanese verbs and cre- 
ated a set of lexical frames, mapping rules, and an inheritance 
hierarchy for use in a working translation system. 
2 Linguistic Motivation 
Our methodology is based in part on recent work in lexical 
semantics (Jackendoff 1983, 1987; Levin, B. 1985, 1987, 
1989; Hale and Keyser 1986; Fukui, Miyagawa, ,and Tenny 
1985; Rappaport and Levin, B. 1986). The field of lexical 
semantics is concerned with therepresentation of syntactically 
relevant aspects of word meaning, especially the properties of 
argument-taking words like verbs. 
Many researchers have noticed that semantically similar 
predicates tend to be syntactically similar, too. B. Levin 
(1987, 1989) examines many systematic semantic-syntactic 
correspondences, including linking regularities and transitiv- 
ity alternations. Linking refers to associations between se- 
mantic arguments and grammatical relations. Common cor- 
respondenees between semantic arguments and grammatical 
relations are called linking regularities. 
2.1 Linking and Alternation 
For example, in the causative use of break (e.g., John broke the 
vase), the subject John is linked to the agent semantic role, and 
the object vase maps to the theme semantic role. Break can 
be classified as a change-of-state verb, and the same pattern is 
observed in the causative use of other change-of-state verbs, 
such as crack and melt. Moreover, it is important to note 
that this pattern also holds for other classes of verbs (e.g., 
change-of-possession verbs like give). 
It is also the case that the same verb can have more than 
one way of linking syntactic functions with semantic roles. 
These different linkings are called valency alternations, which 
include both transitivity alternations and alternate linkings of 
semantic arguments with syntax. 
For example, break can also appear in sentences like The 
vase broke, where the verb assigns the theme semantic role 
to the syntactic subject. This is in contrast to the causative 
use of break, described above, where file verb assigns the 
agent semantic role to the syntactic subject and a theme se- 
ACRES OE COLING-92. N^wrEs, 23-28 ho~r 1992 l 2 5 4 Pgoc. oF COLING-92, NANTES, AUG. 23-28, 1992 
mantic role to the syntactic object. This alternation, known as 
the Causative/lnchoative alternation, is also associated with 
change-of-position verbs like drop (John dropped the ball vs. 
The ball dropped), change-of-psychological-state verbs like 
worry (John worried vs. Bill worried John), etc. (B. Levin, 
1989). 
With some verbs, rite mapping of one syntactic function 
may remain constant while others alternate. For example, 
in the sentence John cut the meat, the patient semantic role 
is assigned to the syntactic object; in John cut at the meat, 
tile goal semantic role is assigned to the prepositional object. 
In both sentences, the agent semantic role is assigned to the 
syntactic subject. 
Classes of verbs which undergo the same alternation tend 
to be semantically similar. Verbs like hack and slash, which 
belong to the same verb class as cut, undergo tile same al- 
ternation mentioned above. However, semantically different 
verbs like break do not exhibit the stone alternation: 
(I) a. He broke the cup. 
b. *He broke at the cup. 
Linking regularities and transitivity alternations are used 
to identify the semantic roles of arguments and the semantic 
classes of verbs. That is, an argument which displays the same 
linking regularities as another argument might be assigned the 
same thematic role, and verbs which have the same transitivity 
alternations can be placed in the same class. 
Transitivity alternations in English are marked in various 
ways. Many of them involve the alternation of an argument 
between object add prepositional phrase. In Japanese, how- 
ever. valency alternations, (including transitivityalternatious) 
are usually indicated by different case markers lx3rne by the 
argmnents of the verb. Every noun phrase in Japanese is 
marked postpositionally by a particle, such as ga, o, ni, and 
de. These markers indicate tile case or other grammatical 
fanction of the nominals they are associated with. 
For example, the o/de alternation appears with verbs like 
oyogu (swim), sanposuru (rake a walk), and hashiru (run)< 
(2) a, Taro ga kawa o oyoida 
~Taro swam down the river' 
b. Taro ga kawa de oyoida 
~Taro swam in the river' 
2.2 Lexical Mapping 
Another part of building a lexical semantic representation is 
to foramlate links from lexical items to conceptual meanings; 
these links are called lcxical mappings. Since the semantic 
properties of relations and objects (which are crucial in stat- 
ing subcategorization restrictions) reside most naturally in a 
semantic domain model, it is necessary for a system to inte- 
grate the lexical level and the domain model so that semantic 
restrictions can be satisfied during parsing and generation. 
In some cases, a lexical item may be linked to more than 
just a semantic head. For exmnple, in the sentence The pen- 
cil rolled off the table, the meaning of roll must be repre- 
i For further detail and examples, see (Mitamura, 1989). 
seated by both a semantic head (e.g., *MOVE) and a seman- 
tic modifier indicating the manner of motion (e.g., (manner 
*ROTATION)) 2. As a result, lexical mapping may also require 
semantic feature assignment. 
2.3 Summary 
The motivation for our work has been the following set of ob- 
servations, drawn from the linguistic phenontena mentioned 
in this section. An appropriate lexical representation must be 
able to represent tile following: 
• The linking of a particuhu" syutactic function with a par- 
ticular semantic role; 
• A set of linking rules that indicate a partianlar alternation; 
• A group of alternations that capture the general behavior 
of a class of verbs; 
• An explicit representation of verb classes, to which par- 
ticular lexical items may be linked; 
. A set of lexical items, which contain both links to verb 
classes and links to semantic concepts in the domain 
conceptual hierarchy. 
3 The Lexical Hierarchy 
Our lexical hierarchy has five levels of representation, each 
corresponding to a linguistically meaningful unit of structure: 
(1) Mapping Rule Frames, which capture a particular corre- 
spondence between a syntactic function and a semantic role; 
(2) Mapping Pattern Fran|es, which capture a particular set 
of mapping rules, which correspond to one way of linking 
the arguments of a particular verb; (3) Mapping "b/pc Frames, 
which capture tile set of alternations (mapping patterns) al- 
luwed by a particular class of verbs; (4) Verb Class Frames. in 
which the generalization in verb linking behavior is captured; 
(5) Lcxical Frames, in which tmrticnlar lexical items (verbs) 
are represented as frames which are linked both to appropri- 
ate verb class frames and to conceptual frames in the donmin 
concept hierarchy. 
Figure 1 illustrates the inheritance relations between map- 
ping rules, mapping patterns, mapping types, verb classes, 
aml lexical frames in English. 
3.1 Mapping Rule Frames 
The nmpping rule frames each map one grammatical function, 
such as subject or object, onto a semantic role, such as agent 
or theme. Each mapping rule is specified in a separate frame, 
as in the lollowing: 
a. ( ~ agent-sub j-mapping 
(:agent subj)) 
b. ( * theme-ob j-mapping 
(:theme obj) ) 
c. ( * theme-sub j-mapping 
(:theme subj) ) 
2This is similar to the nouon of conflation discussed by "litlrny (1985). 
Ac'lxs DE COLING-92, NANTi{S, 23-28 AOUI 1992 I 2 5 5 PROC. OF COLIN(i-92, NANTES. AUO. 23-28, 1992 
3.2 Mapping Pattern Frames 
The mapping pattern frames represent particular bundles of 
mapping rules. For example, a mapping pattern frame which 
contains the agent-subject mapping and the theme-object map- 
ping represents one mapping pattern, whereas a frame which 
contains just the theme-subject mapping represents .another 
mapping pattern 3 (of. Figure 1). 
Syntactic constraint rules can be written in a mapping pat- 
tern frame to indicate that the associated mapping rnles can 
apply only when these constraints are satisfied. Some exam- 
pies of mapping pattern frames are shown below: 
( *mapping-pat ternl 
(syntactic-constraint 
(passive = -) ) 
(contain *theme--ob j-mapping 
* agent- sub j-mapping) ) 
(*mapping-pat tern2 
(syntactic-constraint 
(passive = -) ) 
(contain *theme-subj-mapping) ) 
The frame *mapping-pattern I captures one way of mapping 
the syntactic argument of a verb. The subject is mapped to 
the semantic agent and the object is mapped to the semantic 
theme. The *1napping-pattern2 frame indicates a mapping 
where the verb has one argument, the subject, and maps the 
subject to the semantic agent. 
3.3 Mapping Type Frames 
Mapping type frames contain sets of mapping rule pat- 
terns, and have the ability to capture both transitivity al- 
ternations in English and case alternations iu Japanese (Mi- 
~uura, 1989). The two mapping patterns we mentioned 
earlier, 1) the agent-subject and the theme-object mapping, 
and 2) the theme-subject mapping, can be generalized as the 
causative-inchoative verb mapping type. In Figure 1, the 
causative-inchoative alternation is represented by *causative- 
inchoative, hi Japanese, the alternation between an oblique 
argument with particle o and an oblique argumeat with particle 
de is captured by *obl-o/obl-de. 
An example of a mapping type frame is shown below: 
( *causative-inchoative 
(contain *mapping-pat ternl 
*mapping-pat te rn2 ) ) 
The *causative-incltoative frame contains two mapping pat- 
tern frames, indicated by a contain link that includes 
*mapping-patterul and *mapping-patteru2. 
3This is similar to the notkm of lexicalforms in lexical mappi.g theory 
(nresnan ind Kanerva, 1989), but fire difference is that we incorporate case 
assignment tule~ into argument mspping nlles to make the mapping a o.e 
step operatitx~ for use in generating or par~ing sentences. In LFG, casez are 
tssigned in each lexical entry through grammatical encoding theory, which 
identifies ~t.d assigns an appropriate ease for t grammatical l'unction ill each 
lexieal ealtry. 
3.4 Verb Class Frames 
Verb class frames generalize over verbs with a common core 
sense and common syntactic behavior, Some example verb 
class frames (*verbs-of-breaking, *motion-path-verbs) are il- 
lustrateA in Figure 1. The *verbs-of-breaking frame has an 
is-a link to the *causative-inchoative mapping type, indicat- 
ing that verbs in the *verbs-of-breaking class can undergo the 
causative-inchoative alternation. 
3.5 Lexical Frames 
Lexical frames represent the language-dependent lexicon, and 
include pointers to corresponding conceptual frames. These 
frames also have is-a relations which link them to verb class 
frames, which are organized hierarchically according to the 
particular language, 
The SEMANTICS slot in the lexical frame contains ref- 
ereaces to the concepttml frames associated with the lexical 
item. Particular restrictions on the meaning of the lexical item 
are captured by semantic role or feature assignment rules that 
may appear along with each SEMANTICS pointer. 
For example, the SEMANTICS slot shown below for the 
verb roll points to the conceptual frame *MOVE. Included 
with the pointer to *MOVE is an assignment rule which in- 
dicates that the manner of *MOVE must have the meaning 
indicated by the conceptual frame *ROTATION. The *roll-1 
frame has an is-a relation to the verb class frame, *motion- 
verbs. 
(*roll-i 
{is-a *motion-verbs) 
(semantics 
( * MOVE 
(:manner = *ROTATION) ) ) ) 
More examples of lexical frames are shown in Figures 1. 
I11 Figure 1, *break-1 is a lexical frame, corresponding to the 
semantic notion *BREA'K, which is a member of the *verbs- 
of-breaking verb class. 
4 The Domain Conceptual Hierarchy 
Conceptual frames represent knowledge of the world that is 
language-independent, lbr example, general concepts such as 
• EVENT and *PHYS ICAL-OBJECT, as well as more specific 
concepts, like *BREAK and *SWIM 4. Conceptual frames are 
organized hierarchically using inheritance relations. Selec° 
tional restrictions can be specified in conceptual frames, and 
appear as the fillers of semantic role slots. 
5 Multiple Inheritance and Interpretive 
Mapping in Machine Translation 
Oar operational goals in constructing this hierarchy and its 
inheritance relations include the following: 
'*An asterisk prefix is used to indicate frame names. Upper case frame 
names (e,g,, *BREAK) indicate conceptual frames. Lower ease is used for 
all other frame names (e.g,, lexical frames, veal) class frames, etc.). 
ACTES DE COLING-92, NAtZI'eso 23-28 AOL'T 1992 1 2 5 6 PROC. OF COLING-92, NANTES, AUO, 23-28, 1992 
Mapping (*m- zu).al |*~ rul~ (*a-rt*Im3 ~u~o~ {:,g,~t .ttb:)}) (:tlwm ~9)1 (:tlw,~ ,ub3)) 
conlai~co~ain t ¢°~lain 
k~ap~h30 ( ,l_lmtt~rn I (*m l~tt*=~ Pallems (oontab~ *= xattal (c~ntatn 
*m-ra/.~3) ) 
/ contain ,/1-'"g~c.ontain 
M~PP~O { *~uutlv..-la~eholtiw \[yp~s {o0~tain ~m=i~ttto~nl 
~-~tt,xn2 ) ) 
lisa 
Vopb ( *vlrbl -o £-bzaakin(/ Classos {im-a 
* caul ~t ivo -Inchoat £v4) ) 
l.oxiCal (*br~k-I (*~m~-% Frafr~s (m~ntl~ *14PX.tl) (~mumt£~ *cl~t~() 
Figure 1: Lexieal Hierarchy Example: English 
• Support of rapid, straightforward acquisition of large 
amounts of texical knowledge in an interactive environ~ 
meat; 
• Elimination of unuecessary (and costly) redundancy in 
the representation of lexical knowledge. 
5.1 Efficient Knowledge Aequisitinn 
Productivity in the kuowledge acquisition task is greatly ea- 
hancexl by this hierarchical methodology. Rather than exliting 
an ASCII file containing rednndant mapping rule definitions 
for each lexical entry, the persou entering new lexical con- 
cepts utilizes a 2-dimensional browsing and editing tool to 
add new knowledge to the system (Kaufinann, 1991). 
Once the initial mapping rules, alternations attd verb classes 
are specified, the user can easily link new lexical frames to 
existing verb classes, perhaps refining some of the knowledge 
in tile upper portions of the hierarchy, but in general taking 
advaltlage of the compact uaturc of the hierarchy to avoid 
redund,'mt data entry. 
The fiame representation presented here has a great advtm- 
rage lot the development of large-scale NLP systems, namely, 
that each mapping rifle need only be defined once, anti is there- 
after inherited by all the lexieal frames that require it. By 
positing intermediate levels of structure (mapping types and 
mapping patterns), significant generalizations can be captured 
which farther enhaace the compactness of tile representation 
and the ease of knowledge acquisition. 
5.2 Multiple Inheritance 
The delinition of containment, however, is not as straight- 
lorwm'd as a simple is-a relation in traditional frame-based 
knowledge representation. "lhe containment relation that ob- 
tains between mappingpattems and mapping rules is the usual 
conjunctive (nmltiple) type of inheritance, since a mapping 
pattern contains each and every mapping rule that it is linked 
to via a contain link. On the other hand, the containment rela- 
tion that holds between mapping types and mapping patterns 
in disjunctive, since a nmpping tyve contains different types 
of alteraations, only one of which can be active at a given time 
for a particular verb. As a result, inheritance is pert'ormed in 
a different manner at these two levels in the hierarchy. 
By default, FrameKit supports only conjunctive inheri- 
lance, which is most common in system where inherit.mtee 
hierarchies are built using simple is-a links. We have de- 
veloped user-delined inheritance methods for FrameKit that 
perform rite appropriate inheritance operations at each level 
in the mapping Itierarchy. When all of the possible subcate- 
gorization/nmpping pairs must be retrieved for a given lexi- 
cal frame, these inheritance methods perform the appropriate 
conjunctive inheritance, bundliug the mapping rules together 
into mapping patterns, followed by disjunctive inheritance of 
mapping types to create any "alternative readings of the lexical 
item. Simply speaking, the inheritance methods must re- 
create the explicit structure that is implicit in the inheritance 
hierarchy when it is necessary to represent distinct mappings 
for verbs at system run-time. 
An extuuple of how iuberitance works at run time is illus- 
trated in Figure 2. The two franles shown in the figure are 
instantiated by the inheritance methods from the lexical frame 
*break-l, and represent the two possible alternations of break 
(the causative reading and file inchoative reading). 
( * BF, EAK- 1153 
(TtlEME OBJ) 
(AGENT SUBJ) 
(SEMANTICS *BREAK) 
(CREATED-FROM *BREAK-I) ) 
(*BREAK-If 54 
(THEME SUBJ) 
(SEMANTICS *BREAK) 
(CREATF, D-FROM *BREAK-I ) ) 
Figure 2: lnstantiated Frames for *break-1 
5.3 Interpretive Mapping 
The architecture in Figure 3 illustrates how our lexieal hier- 
archy fits into the overall machine translation system. During 
prosing, tile lexieal entries stored iu the source lexical hievaro 
chy are accessed by the LF(I parser; during the mapping of 
source f-stractures to interlinguarepresentations, file mapping 
rules in the lexical hierarchy are accessed by the mapper via 
instarlliated mapping structures like those shown in Figure 
2. During generation, tile target language lexical hierarchy 
is utilized in a similar fashion. First, instantiated ntapping 
structures arc used to create target f-structures, a,d then tar- 
get lexieal entries are utilizexl by the LFG generator to produce 
target language strings. 
6 Status 
we have developed an extensive interpretive mapping hier- 
archy for Japzmese, which includes 36 mapping rule frames, 
Acrus DI~ COLING-92, NAtal,S. 23-28 AOUT 1992 I 2 5 7 PRoc. OF COL1NG-92, NANTI~S. AUG. 23-28, 1992 
Data Procnelng Knowledge 
Flow Modules Sources 
Figure 3: System Architecture for Machine Translation 
45 mapping pattern frames, 37 mapping type frames, 54 verb 
class frames, and 100 lexical frames. Hundreds of additional 
lexical frames could be added to the hierarchy without mod- 
ification of the existing hierarchical structure. We believe 
that our mapping frame hierarchy accounts for the syntactic 
behavior of a significant number of Japanese verb classes. 
The hierarchy is based on data for about 1000 verbs, taken 
from (Ishiwata and Ogino, 1983) and the IPAL report on basic 
Japanese verbs OPAL, 1987), 
We have also developed an initial mapping hierarchy for 
English verbs. The English and Japanese lexical hierarchies 
were utilized in the KBMT-89 system for the interpretation of 
Japanese sentences (Mitamura, et al., 1991). Weare currently 
integrating our hierarchical structure into a large-scale system 
for translation of service manuals from English to Japanese. 
Since the argument mapping knowledge represented in our 
hierarchy is declarative rather than procedural, it can be used 
either in analysis or generation (cf. Figure 3). 
7 Conclusion 
High-quality machine translation requires an adequate seman- 
tic interpretation of the source text. To achieve this goal, we 
feel it is necessary to incorporate the kind oflexical knowledge 
and structure that has been explored in the theory of lexical 
semantics. We have presented a methodology that can be used 
to construct lexical hierarchies which represent lexical knowl- 
edge in a compact, efficient representation which captures rel- 
evant linguistic generalizations, as well as providing a useful 
framework for knowledge acquisition and system-building. 
This methodology is declarative and language-independent, 
and can be used either for parsing or generation. 
References 
\[1\] Bresnan. J. and J. M. Kanerva (1989) "Locative Inversion 
in Chichewa: A Case Study of Factorization in Grammar," 
Linguistic Inquiry, 20:1.1-50. 
\[2\] Carbonell, J. G. and M. Tomita (1987). "Knowledge-based 
Machine Translation: The CMU Approach," in Nirenburg, S. 
(ed,), Machine Translation: Theoretical and Methodological 
Issues, New York: Cambridge University Press. 
\[3\] Fukui. N., S. Miyagawa. and C. Tenny (1985) "Verb Classes 
in English and Japanese: A Case Study in the Interaction of 
Syntax, Morphology and Semantics,"LexiconProject Working 
Papers #3, Center for Cognitive Science, MIT, Cambridge, 
MA. 
\[4\] Goodman, K. and S. Nirenburg, eds. (1991) A Case Study 
in Knowledge-Based Machine Translation, San Mateo, CA: 
Morgan Kanfmann. 
\[5\] Hale, K. and J. Keyser (1986) "SomeTransitivity Altematinns 
in English," Lexicon Project Working Papers #7, Center for 
Cognitive Science, MIT, Cambridge, MA. 
\[6\] Ishiwata, T. and T. Ogino (1983) "Nihongo Yougen no Ket- 
sugoka," Bunpou to Imi L Asakura-shoten, Tokyo. 226-272. 
\[7\] Jackendoff. R.S. (1983) Semantics and Cognition, MIT Press, 
Cambridge. MA. 
\[8\] Jackendoff, R.S. (1987)''The Status of Thematic Relations in 
Linguistic Theory," Linguistic Inquiry, 18:3,369-411. 
I9\] Johoushori shinkou jigyou kyoukai, ed. (1987) Keisanki-you 
Nihongo Kihon-doushi Jisho IPAL - Basic Verbs: jisho-hen. 
\[10\] Kanfmann. T. (1991). The ON'I"OS User's Guide. Technical 
Mmno, Center for Machine Translation, Carnegie Mellon Uni- 
versity. 
\[11\] Levin, It. (1985) "Lexical Semantics in Review: an Introduc- 
tion." Lexicon Project Working Papers #1, Center for Cogni- 
tive Science, MFF, Cambridge, MA. 
\[12\] Levin, B. (1987)"Approaches to Lexical Semantic Represen- 
tation," Unpublished Manuscript. 
\[13\] Levin, B. (1989) "English Verb Diathesis," Lexicon Project 
Working Papers#32, Center for Cognitive Science, MIT, Cam- 
bridge, MA. 
\[14\] Mitamura, T. (1989) "The Hierarchical Organization of Pred- 
icate Frames for Interpretive Mapping in Natural Language 
Processing," Ph.D. dissertation, University of Pittsburgh. 
\[15\] Mitamura, T., et al. (1991) "Analysis Lexicon," in Goodman 
and Nirenburg, eds., A Case Study in Knowledge-Based Ma- 
chine Translation, San Mateo, CA: Morgan Kaufmann. 
\[16\] Nyberg, E. (1988) The FrameKit User's Guide: Version 2.0, 
Technical Report, Center for Machine Translation, Carnegie 
Mellon University, CMU-CMT-88-MEMO. 
\[17\] Rappaport, M. and B. Levin (1986) "'What to Do with Theta- 
Roles," Lexicon Project Working Papers #11, Center for Cog- 
nitive Science, M1T. Cambridge, MA. 
\[18\] Talmy, L. (1985) "Lexicalization l'anems: Semantic Structure 
in Lexical Forms," in T. Shopen, ed.. Grammatical Categories 
and the Lexicon, Language Typology and Syntactic Descrip- 
tion 3, Cambridge University Press. 
ACTES DE COLING-92, NAN'rES, 23-28 AOI~T 1992 1 2 5 8 PROC. OF COLING-92, NANTES, AUG. 23-28, 1992 
