On UNL as the future "html of the linguistic content" & the reuse of 
existing NLP components in UNL-related applications with the 
example of a UNL-French deconverter 
Gilles St~RASSET 
GETA, CLIPS, IMAG 
385, av. de la biblioth~que, BP 53 
F-38041 Grenoble cedex 9, France 
Gilles.Serasset@ imag,fr 
Christian BOITET 
GETA, CLIPS, IMAG 
385, av. de la biblioth~que, BP 53 
F-38041 Grenoble cedex 9, France 
Christian.Boitet @imag.fr 
Abstract 
After 3 years of specifying the UNL (Universal Networking Language)  and 
prototyping deconverters I from more than 12 s and enconverters for about 4, the 
UNL project has opened to the community by publishing the specifcations (v2.0) of the UNL 
, intended to encode the meaning of NL utterances as semantic hypergraphs and to be 
used as a "pivot" representation in multilingual information and communication systems. 
A UNL document is an html document with special tags to delimit the utterances and their 
rendering in UNL and in all natural s currently handled. UNL can be viewed as the 
future "html of the linguistic content". It is only an interface format, leading as well to the reuse 
of existing NLP components as to the development of original tools in a variety of possible 
applications, from automatic rough enconversion for information retrieval and information 
gathering translation to partially interactive enconversion or deconversion for higher quality. 
We illustrate these points by describing an UNL-French deconverter organized as a specific 
"localizer" followed by a classical MT transfer and an existing generator. 
Keywords 
UNL, interlingua, pivot, deconversion, UNL~French localization, transfer, generation. 
Introduction 
The UNL project of network-oriented 
multilinguat communication has proposed a 
standard for encoding the meaning of natural 
 utterances as semantic hypergraphs 
intended to be used as pivots in multilingual 
information and communication systems. In the 
first phase (1997-1999), more than 16 partners 
representing 14 s have worked to build 
deconverters transforming an (interlingual) 
UNL hypergraph into a natural  
utterance. 
In this project, the strategy used to achieve this 
initial objective is free. The UNL-French 
deconverter under development first performs a 
"localization" operation within the UNL format, 
and then classical transfer and generation steps, 
using the Ariane-G5 environment and some 
UNL-specifc tools. 
The use of classical transfer and generation 
steps in the context of an interlingual project 
may sound surprising. But it reflects many 
interesting issues about the status of the UNL 
, designed as an interlingua, but 
diversely used as a linguistic pivot (disambi- 
guated abstract English), or as a purely semantic 
pivot. 
After introducing the UNL , we present 
the architecture of the UNL-French deconverter, 
which "generates" from the UNL interlingua by 
first "localizing" the UNL form for French, 
within UNL, and then applying slightly adapted 
but classical transfer and generation techniques, 
implemented in the Ariane-G5 environlnent, 
supplemented by some UNL-specific tools. 
Then, we discuss the use of the UNL  
as a linguistic or semantic pivot for highly 
multilingual information systems. 
1 The UNL project and  
1.1 The project 
UNL is a project of multilingual personal 
networking communication initiated by the 
University of United Nations based in Tokyo. 
The pivot paradigm is used: the representation 
I The terms << deconvcrsion, and <~ enconvcrsion, are specific to tile UNL proiect and are defined at paragraph 2. 
768 
of an utterance in the UNL interlingua (UNL 
stands for "Universal Networking Language") is 
a hyl)ergraph where normal nodes bear UWs 
CUniversal Words", or interlingual acceptions) 
with semantic attributes, and arcs bear semantic 
relations (deep cases, such as agt, obj, goal, etc.). 
Hypernodes group a subgraph defined by a set 
of connected arcs. A UW denotes a set of 
interlingual acceptions (word senses), although 
we often loosely speak of "the" word sense 
demoted by a UW. 
Because English is known by all UNL 
developers, the syntax of a tlormal WW is: 
"<English word or compound> ( <list 
of restrictions> ) ", O. Z. "look for 
(icl>action, agt>human, obj>thing)" 
Going fronl a text to the corresponding "UNL 
text" or interactively constructing a UNL text is 
called "enconversioif', while producing a text 
fiom a sequence of UNL graphs is called 
"deconversion". 
This departure fi'om the standard terms of 
analysis and generation is used to stress that this 
is not a classical M\]: projecl, bu! that UNL is 
planned to be the source format preferred for 
representing textual inl:ormation in tile 
envisaged multilingual network environment. 
Tile schedule of tile project, beginning with 
deconversion rather than cnconvcrsion, also 
reflects that difference. 
14 hmguages have been tackled during the first 
3--year phase of the prqject (1997-1999), while 
many more arc to be added in tile second 
phase. Each group is fi-ee to reuse its own 
software lools and/or lingware resources, or to 
develop directly with tools provided by tile 
UNL Center (UNU/IAS). 
Emphasis is on a very large lexical coverage, so 
that all groups spend most of their time on tile 
UNL-NL lexicons, and develop tools and 
methods for efficient lexical development. By 
contrast, gramnmrs have been initially limited to 
those necessary for deconversion, and will then 
bc gradually expanded to allow for more 
naturalness m formulating text to be 
enconverted. 
1.2 The UNL components 
1.2.1 Universal Words 
Tile nodes of a UNL utterance are called 
Universal Words (or Uws). The syntax of a 
normal UW consists of 2 parts : 
a headword, 
a list of restrictions 
Because English is known by all UNL 
developers, tile headword is an English word or 
compound. The restrictions are given as all 
attribute value pail" where attributes are semantic 
relation labels (as the ones used in the graphs) 
and wllues are other UWs (restricted or not). 
A UW denotes a collection of interlingual 
acceptions (word senses), although we often 
loosely speak of "the" word sense denoted by an 
UW. For example, the unrestricted UW "look 
for" denotes all the word-senses associated to 
tile English compound word "look for". Tile 
restricted UW " look for ( icl>action, 
agt>human, obj>thing) " represents all tile 
word senses of the English word "look for" 
that are an action, perl%rmed by a human that 
affects a thing. In this case this leads to the word 
sense: "look for- to try to find". 
1.2.2 UNL hypergraph 
A UNL expression is a hypergraph (a graph 
where a node is simple or recursively contains a 
hypergraph). Tile arcs bear semantic relation 
labels (deep cases, such as agt, obj, goal, etc.). 
score(icl>event agt>human,tld>sport) \[ I @ entry. @ past. @ complete \[ 
7 i \ 
agt ....... '/ ins \~ 
i Rona~do 1 °b~ / \ pit head(p~ol>body) ~ "\\ 
/ ,~ corner 
/ 
goal i~cl>thing) \[.41 obj mod \[ 
~left : 
Figm'e I. 1: A UNL graph deconvertible as "Ronaldo 
has headed the ball into the left corner of the net" 
In a UNL graph, UWs appear with attributes 
describing what is said from tile speaker's point 
of view. This includes phenomena like speech 
acts, truth wllues, time, etc. 
Hypernodes may also be used ill UNL 
expressions. 
agt ...... i 
I driver.~Pl \] 
aoj 
I reckless \] 
01.@entry 
\[ drink\] 
\ drive \] 
Figure 1.2: A UNL Io,pergraph that may be 
deconverted as "Reckless drivers drink and drive" 
Graphs and subgraphs nmst contain one special 
node, called the entry of tile graph. 
1.2.3 Denoting a UNL graph 
These hypergraphs are denoted using the UNL 
 per se. In the UNL hmguagc, an 
769 
expression consists in a set of arcs, connecting 
the different nodes. As an example, the graph 
presented in figure 1.1 will be denoted as: 
agt(score(...).@entry.@past.@complete, 
Ronaldo) 
obj(score(_.).@entry.@past.@complete, 
goal(icl>thing)) 
ins(score(...) .@entry.@past.@complete, 
head(pof>body)) 
plt(score(...) .@entry.@past.@complete, 
corner) 
obj (corner, goal(icl>thing)) 
mod(corner, left) 
Hypernodes are denoted by numbers. The 
graph contained by a hypernode is denoted as a 
set of arcs colored by this number as in: 
agt (:Ol.@entry, driver. @pl) 
aoj (reckless, driver.@pl) 
and:Ol (drive, drink.@entry) 
Entries of the graph and subgraphs are denoted 
with the ".@entry" attribute. 
2 Inside the French deconverter 
2.1 Overview 
Deconversion is the process of transforming a 
UNL graph into one (or possibly several) 
utterance in a natural . Any means 
may be used to achieve this task. Many UNL 
project partners use a specialized tool called 
DeCo but, like several other partners, we choose 
to use our own tools for this purpose. 
One reason is that DeCo realizes the 
deconversion in one step, as in some transfer- 
based MT systems such as METAL \[17\]. We 
prefer to use a more modular architecture and 
to split deconversion into 2 steps, transfer and 
generation, each divided into several phases, 
most of them written in Arlene-G5. 
Another reason for not using DeCo is that it is 
not well suited for the morphological gene- 
ration of inflected s (several thousands 
rules are needed for Italian, tens of thousands 
for Russian, but only about 20 rules and 350 
affixes suffice to build an exhaustive GM for 
French in Sygmor). Last, but not least, this 
choice allows us to reuse modules already 
developed for French generation. 
This strategy is illustrated by figure 2.1. 
/~;~,; o_;.,,.~,', Transfer 
~ ...... v" Ge I~ati0n her 
/ \,4v 
French utterance 
Fig. 2.1:2 possible deconversqon strategies 
Using this approach, we segment the decon- 
version process into 7 phases, as illustrated by 
figure 2.2. 
The third phase (graph-to-tree) produces a 
decorated tree which is fed into an Ariane-G5 
TS (structural transfer). 
Valklatiolff l,exicaI l'lansl~.'r (h~,li~h 1o tree \[ .ocalization COllversion 
,Z~ "UNL Tree" 
l'araphra~c choice 
UMA structure?N\ 
Syntactic ~gcnerali(ln ,t 
UMC structure~ 
Morl~lml.gic\[ll generation 't 
French utterance 
Fig. 2.2: architecture of the French deconverter 
2.2 Transfer 
2.2.1 Validation 
When we receive a UNL Graph for decon- 
version, we first check it for correctness. A UNL 
graph has to be connected, and the different 
features handled by the nodes have to be 
defined in UNL. 
If the graph proves incorrect, an explicit error 
message is sent back. This validation has to be 
performed to ilaprove robustness of the 
deconverter, as there is no hypothesis on the 
way a graph is created. When a graph proves 
valid, it is accepted for deconversion. 
2.2.2 Loeal&ation 
In order to be correctly deconverted, tile graph 
has to be slightly modified. 
2.2.2.1 Lexical localization 
Some lexical units used in the graph may not be 
present in the French deconversion dictionary. 
This problem may appear under different 
circumstances. First, the French dictionary 
(which is still under development) may be 
incomplete. Second, the UW nmy use an 
unknown notation to represent a known French 
word sense, and third, the LAV may represent a 
non-French word sense. 
We solve these problems with the same method : 
Let w be a UWin the graph G. Let D be the 
French dictionary (a set of UWs). We substitute 
w in G by w' such that: w' e D and 
VxeD d(w, w', G) = d(w, x, G). where d is a 
pseudo-distance function. 
770 
If different French UWs are at the same pseudo- 
distance of w, w' is chosen at random among 
these UWs (default in non-interactive mode). 
2.2.2.2 "Cultural" localization 
Some crucial information may be missing, 
depending on the  of the source 
utterance (sex, modality, number, determination, 
politeness, kinship...). 
It is in general impossible to solve this problem 
fully automatically in a perfect manner, as we 
do not know anything about the document, its 
c:ontext, and its intended usage: FAHQDC 2 is no 
more possible than FAHQMT on arbitrary texts. 
We have to rely on necessarily imperfect 
heuristics. 
ttowever, we can specialize tile general French 
deconverter to produce specialized servers for 
different tasks and different (target) 
subs. It is possible to assign priorities 
not only to various parts of the dictionaries 
(e.g., specialized vs. general), but also to 
equivalents of the same UW within a given 
dictionary. We can then define several user 
profiles. It is also possible to build a memory of 
deconverted and possibly postedited utterances 
for each specialized French deconversion 
server. 
2.2.3 Lexical Transfer 
After the localization phase, we have to perform 
the lexical transfer. It would seem natural to do 
ill within Ariane-G5, after converting the graph 
into a tree. But lexical transfer is context- 
sensitive, and we want to avoid the possibility of 
transferring differently two tree nodes 
corresponding to one and the same graph node. 
Each graph node is replaced by a French lcxical 
unit (LU), along with some variables. A lexical 
unit used in tile French dictionary denotes a 
derivational family (e.g. in English: destroy 
denotes destroy, destruction, destructible, 
destructive .... in French: d6truire for d6truire, 
destruction, destructible, indestructible, 
destructif, destructeur). 
There may be several possible lexical units for 
one UW. This happens when there is a real 
synonymy or when different terms are used in 
different domains to denote the same word 
sense 3. In that case, we currently choose tile 
lexical unit at random as we do not have any 
information on tile task the deconverter is used 
for. 
Tile same problem also appears because of tile 
slrategy used to build the French dictionary. In 
order to obtain a good coverage from the 
beginning, we have underspecified tile UWs and 
linked them to dift'ercnt lexical units. This way, 
we considered a UW as tile denotation of a set 
of word senses in French. 
Hence, we were able to reuse previous 
dictionaries and we can use the dictionary even 
if it is still under development and incolnplete. 
In our first version, we also solve this problem 
by a random selection of a lexical unit. 
2.2.4 Graph to tree conversion 
The subsequent deconversion phases are 
performed in Ariane-G5. Hence, it is necessary 
to convert the UNL hypergraph into an Ariane- 
G5 decorated tree. 
The UNL graph is directed. Each arc is labelled 
by a semantic relation (agt, obj, ben, con...) and 
each node is decorated by a UW and a set of 
features, or is a hypernode. One node is 
distinguished as the "entry" of the graph. 
An ARIANE tree is a general (non binary) tree 
with decorations on its nodes. Each decoration 
is a set of wlriable-value pairs. 
The graph-to-tree conversion algorithln has to 
lnaintain the direction and labelling of the 
graph along with the decoration ot' the nodes. 
Our algorithm splits tile nodes that are the target 
of more than one arc, and reverses the direction 
of as few arcs as possible. An example of such a 
conversion is shown in figure 2.3. 
! a \] 
\[5E3 /\ 
J x, 
x y 
I~! Ice 
z t 
b (x 
I I 
=> l 
d : z +~ 
c:Y i 
¢:1 
Fig. 2.3: example graph to tree convel:vion 
Let Z be the set of nodes of G, A the set of 
labels, T the created tree, and N is the set of 
nodes of T. 
Tile graph G={ (a,b,l) lac Y.,b6 Z,I~ A} is 
defined as a set of directed labelled arcs. We use 
an association list A = { (n,;,n.r) I ,,,+ ~ r,, U. r E 
N }, where we memorize the correspondence 
between nodes of the tree and nodes of the 
graph. 
2 fully autonmtic high quality dcconvcrsion. 
3 strictly speaking, tile same collection of intcrlingual 
woM senses (acccptions). 
771 
let e(; e   such that e is the entry of G 
e r 6- new tree-node (ed, entry) 
inT +- er(); N 6- {e,r\]; A <-- {(ec;,eT)} 
while G :~ O do 
if there is (a,b,l) in G such that 
G ~- G\(a,b,l); 
b r 6- new tree-node(b, i) ; 
A 6- A <J {(b,b,,)); 
let a, r e N such that (a,a, r) e A 
in add b r to the daughters of a,r; 
else if there is (a,b,l) in G such that (b,br) 6 
G e- G\(a,b,l); 
a T (-new tree-node(a,l i); 
A <--- A U {(a,a.r)}; 
let brl,e N such that (b,br) e A 
in add a,, to the daughters of br; 
else exit on error ("non connected graph"); 
(a, a. r) e A then 
A then 
2.2.5 Structural transfer 
The purpose of the structural transfer is to 
transform the tree obtained so far into a 
Generating Multilevel Abstract (GMA) structure 
\[4\]. 
In this structure, non-interlingual linguistic 
levels (syntactic functions, syntagmatic 
categories...) are underspecified, and (if 
present), are used only as a set of hints for the 
generation stage. 
2.3 Generation 
2.3.1 Paraphrase choice 
The next phase is in charge of the paraphrase 
choice. During this phase, decisions are taken 
regarding the derivation applied to each lexical 
unit in order to obtain the correct syntaglnatic 
category for each node. During this phase, the 
order of appearance and the syntactic functions 
of each parts of the utterance is also decided. 
The resulting structure is called Unique 
Multilevel Abstract (UMA) structure. 
2.3.2 Syntactic and morphological generation 
The UMA structure is still lacking the syntactic 
sugar used in French to realize the choices 
made in the previous phase by generating 
articles, auxiliaries, and non connected 
compunds such as ne...pas, etc. 
The role of this phase is to create a Unique 
Multilevel Concrete (UMC) structure. By 
concrete, we mean that the structure ~s 
projective, hence the corresponding French text 
may be obtained by a standard left to right 
traversal of the leaves and simple morphological 
and graphemic rules. The result of these phases 
is a surface French utterance. 
3 Different uses of the UNL  
3.1 Hypergraphs vs colored graphs 
As presented in section 1.2.3, the syntax of the 
UNL  is based on the description of a 
graph, arc by arc. Some of these arcs are 
"coloured" by a number. This colouring is 
currently interpreted as hypernodes (nodes 
containing a graph, rather than a classical UW). 
This interpretation is arbitrary and imposes 
semantic constraints on a UNL utterance: 
the subgraph (the set of arcs labeled with 
the same colour) is connected, 
arcs with different colours cannot be 
connected to the same node. 
However, even if one uses the UNL  
for a particular kind of application, a different 
interpretation may be chosen. By adding new 
semantic constraints to UNL expressions, one 
may restrict to the use of trees. On the contrary, 
by loosening semantic constraint, one may use 
colored graphs instead of the more restrictive 
hypergraphs. 
This flexibility of UNL may lead to uses that 
differ from the computer science point of view 
(different structures leading to different kinds 
of methods and applications) as well as from the 
linguistic point of view (different ways to 
represent the linguistic content of a utterance). 
This kind of structure is very useful to represent 
some utterances like "Christian pulls Gilles' 
leg". Using a colored graph, one can represent 
the utterance with the graph shown in figure 
3.1, which is not a hypergraph. 
772 
01 .@entry ag t.. 
i\[ pull.@entry i 
\[ Chns~lian \] I ~,obj 
pos 
G es \] 
Figure 3.1: this graph is not cut hypergral)h, it can 
however be represented in UNL htnguage 
When using normal hypergraphs, one could 
only represent the utterance as shown in figure 
3.2. 
agt .... \[ make fun of i 
i Chns'~tan I , i obj 
i i 
Figure 3.2: this graph is a valid hyperglztph 
Heuce, keeping backward compatibility with 
other UNL based systems, one may develop an 
entirely new and more powerfld kind of 
application. 
3.2 Linguistic vs senmntie pivot 
The UNL  defines the interface 
structure to be used by applications (either a 
hypergraph or a colored graph). However, it 
does not restrict the choice of the data to be 
encoded. 
Since tile beginning, two possible and wflid 
apl~roaches has been mentioned. During the 
kickoff meeting of tile UNL prelect, Pr. Tsujii 
prolnoted the use of UNL as a linguistic pivot. 
With this approach, a UNL utterance should be 
the encoding of the deep structure of a valid 
English utterance that reflects the meaning of 
the source utterance. With this approach, the 
German sentence "Hans schwimt sehr gern" 
should be encoded as shown in figure 3.3. 
agt.. _ - like.@entry ~. .. 
\[ Ha-'~s \[ ' "-. man I ob j "-. "A, 
"~--agt ........ \[ s~wim \] i much , 
Figmv 3.3: a linguistic encoding of "ltcms schwimt 
sehr gern " 
On the opposite, Hiroshi Uchida promotes the 
use of UNL as a semantic pivot. With this 
second approach, the same sentence should be 
encoded as shown in figure 3.4. 
agt /zswim.l@entry 
/// I ~ iman 
H wil~lgly 
lined 
Figure 3.4: a semantic encoding of "ltans schwimt sehr 
gem" 
Each approach has its advantages and 
drawbacks and the choice between them can 
only be made with an application in mind. The 
linguistic approach leads to a better quality ill 
the produced results and is an answer to highly 
multilingual machine translation projects. With 
this approach, the UNL graphs can only be 
produced by people mastering English or by 
(partially) automatic enconverters. 
With the semantic approach, subtle differences 
in source utterances (indefinite, reflexivity...) 
can not be expressed, leading to a lower quality. 
However, using this approach, the UNL 
encoding is much more natural and easy to 
perform by a non English speaker (as the 
semantic relations and UWs are expressed at the 
source level). Hence, this approach is to be used 
for multilingual casual communication where 
users may express themselves by directly 
encoding UNL expressions with an appropriate 
editing tool. 
Conclusion 
Working oil tile French deconvel-ter has led to 
im interestiug architecture where deconversion, 
in principle a "generation from interlingua", is 
implemented as transfer + generation from all 
abstract structure (UNL hypergraph) produced 
from a NL utterance. The idea to use UNL for 
directly creating documents gets here an 
indirect and perhaps paradoxical support, 
although it is clear that considerable progress 
and innovative interface design will be needed 
to make it practical. 
However, the UNL  proves flexible 
enough to be used by very different proiects. 
Moreover, with deconverters currently 
developed for 14 s, joining the UNL 
project is really attractive. Let's hope that this 
effort will help breaking the  barriers. 
Acknowledgements 
We would like to thank the sponsors of the UNL 
project, especially UNU/IAS (T. Della Senta) & 
ASCII (K.Nishi) and of the UNL-FR subproject, 
especially UJF (C. Feuerstein), IMAG 
(J. Voiron), CLIPS (Y. Chiaramella), and the 
773 
French Ministery of Foreign Affairs (Ph. Perez), 
as well as the members of UNL Center, 
especially project leader H. Uchida, M. L. Zhu, 
and K. Sakai. Last but not least, other members 
of GETA have contributed in many ways to the 
research reported here, in particular N. N6deau, 
E. Blanc, M. Mangeot, J. Sitko, L. Fischer, 
M. Tomokiyo, and K. Fort. 

References 

\[1\] Blanc l~. & Guillaume P. (1997) 
Developing MT lingware through hlternet : 
ARIANE and the CASH interface. Proc. Pacific 
Association for Computational Linguistics 1997 
Conference (PACLING'97), Ohme, Japon, 2-5 
September 1997, vol. 1/1, pp. 15-22. 

\[2\] Blanehon H. (1994) Persl)ectives of DBMT 
for monolingual authors on the basis of LIDIA-I, an 
implemented mockup. Proc. 15th International 
Conference on Computational Linguistics, 
COLING-94, 5-9 Aug. 1994, vol. 1/2, pp. 
115--119. 

\[3\] Boitet C., R6d. (1982) "DSE-I"-- Le point 
sur ARIANE-78 ddbut 1982. Contrat ADI/CAP- 
Sogcti/Champollion (3 vol.), GETA, Grenoble, 
fdvrier 1982, 400 p. 

\[4\] Boitet C. (1994) Dialogue-Based MT attd 
se(f exl)lahting documents as atl alternative to 
MAHT and MT of controlled s. Proc. 
Machine Translation 10 Years On, 11 - 14 Nov. 1994, 
Cranfield University Press, pp. 22.1--9. 

\[5\] Boitet C. (1997) GETA's MT methodology 
attd its current development towards petwonal 
networking communication attd speech translation in 
the context of the UNL and C-STAR projects. Proc. 
PACLING-97, Ohme, 2-5 September 1997, Meisei 
University, pp. 23-57. (invited communication) 

\[6\] Boitet C. & Blanehon H. (1994) 
Multilingual Dialogue-Based MT for monolingual 
authotw: the LIDIA project arm a fil:s't mockup. 
Machine Translation, 9/2, pp. 99--132. 

\[7\] Boitet C., Guillaume P. & Qu6zel- 
Ambrunaz M. (1982) ARIANE-78, an 
hltegrated environment for atttomated translation attd 
human revision. Proc. COL1NG-82, Prague, July 
1982, pp. 19--27. 

\[18\] Brown R. D. (1989) Augmentation. 
Machine Translation, 4, pp. 1299-1347. 

\[19\] Ducrot J.-M. (1982) TITUS IV. In 
"Information research in Europe. Proc. ot' tile 
EURIM 5 conf. (Versailles)", P. J. Taylor, cd., 
ASLIB, London. 

\[10\] Kay M. (1973) The MIND system. In 
"Courant Computer Science Symposium 8: Natural 
Language Processing", R. Rustin, ed., Algorithmics 
Press, Inc., New York, pp. 155-188. 

\[11\] Maruyama H., Watanabe H. & Oglno S. 
(1990) An Interactive Japanese Parser for Machine 
Translation. Proc. COLING-90, Helsinki, 20-25 
aofit 1990, ACL, vol. 2/3, pp. 257-262. 

\[12\] Melby A. K., Smith M. R. & Peterson 
J. (1980) ITS : An Interactive Translation 
System. Proc. COLING-80, Tokyo, 30/9-4/10/80, 
pp. 424---429. 

1113\] Moneinme W. (1989) (159 p. 
+annexes)7;40 vet's" l'arabe. Sp&'ification d'une 
gdudration sfatldard de l'arabe. Rdalisation d'un 
l)romO'l)e anglais'-ambe ?t partir d'un attalyseur 
existant. Nouvelle thbse, UJF. 

\[114\] Nirenburg S. & al. (1989) KBMT-89 
Project Report. Center for Machine Translation, 
Carnegie Mellon University, Pittsburg, April 1989, 
286 p. 

\[115\] Nyberg E. H. & Mitamura T. (1992) 
The KANT system: Fast, Accurate, High-Quality 
Translation in Practical Domains. Proc. COLING- 
92, Nantes, 23-28 July 92, ACL, vol. 3/4, pp. 
1069--1073. 

\[16\] Qu6zel-Ambrunaz M. (1990) Ariane-G5 
v.3 - Le moniteut: GETA, IMAG, juin 1990, 206 p. 

\[17\] Sloeum J. (1984) METAL: the LRC 
Machine Translation O,stem. In "Machine 
Translation today: the state of the art (Proc. third 
Lugano Tutorial, 2-7 April 1984)", M. King, cd., 
Edinburgh University Press (1987). 

\[18\] Wehrli E. (1992) The IPS System. Proc. 
COLING-92, Nantes, 23-28 July 1992, vol. 3/4, pp. 
870-874.  
