Proceedings of the Workshop on Multilingual Language Resources and Interoperability, pages 32–39,
Sydney, July 2006. c©2006 Association for Computational Linguistics
The Development of a Multilingual Collocation Dictionary 
 
 
Sylviane Cardey 
Centre Tesnière 
Université de Franche-Comté 
France 
sylviane.cardey@univ-
fcomte.fr
Rosita Chan 
Centre Tesnière 
& 
University of Panama 
chan.rosita@hot
mail.com
Peter Greenfield 
Centre Tesnière 
Université de Franche-Comté 
France 
peter.greenfield@univ-
fcomte.fr
 
 
 
Abstract 
In this paper we discuss the development 
of a multilingual collocation dictionary 
for translation purposes. By ‘collocation’ 
we mean not only set or fixed expres-
sions including idioms, simple co-
occurrences of items and metaphorical 
uses, but also translators’ paraphrases. 
We approach this problem from two di-
rections. Firstly we identify certain lin-
guistic phenomena and lexicographical 
requirements that need to be respected in 
the development of such dictionaries. 
The second and other direction concerns 
the development of such dictionaries in 
which linguistic phenomena and lexico-
graphic attributes are themselves a means 
of access to the collocations. The linguis-
tic phenomena and lexicographical re-
quirements concern variously placing the 
sense of collocations rather than head-
words or other access methods at the cen-
tre of interest, together with collocation 
synonymy and translation equivalence, 
polysemy and non-reversibility of the 
lexis, and other more lexicographic prop-
erties such as varieties of language and 
regionalisms, and types of translation. 
1 Introduction 
In work with developing multilingual collocation 
based dictionaries for translation purposes across 
a wide variety of domains (Cardey and 
Greenfield, 1999; Chan 2005) various interesting 
linguistic phenomena and lexicographic require-
ments have been observed. In the context of such 
dictionaries, by the term collocation we include 
not only set or fixed expressions (Moon, 1995, 
Tables 1.1 and 1.2, pp.19-20) including idioms, 
simple co-occurrences of items (plane ticket) and 
metaphorical uses (spill the beans), but also, as 
we will show, translators’ paraphrases where 
these are needed. Linguistic phenomena include 
ones concerning sense (for example synonymy 
and translation equivalence, polysemy and non-
reversibility). Lexicographical requirements in-
clude for example the requirement (for consis-
tency purposes amongst others) that the colloca-
tion (as article) be the centre of interest rather 
than the headword(s) whose role is one of access 
to the collocations. This is principally because 
the object of such dictionaries should be based 
on inter-lingual collocation sense group corre-
spondence, translation of headwords being essen-
tially incidental. Another way to view this is that 
if what we wish to model is a dictionary of 
senses, these senses are expressed by interpreta-
tions in the form of collocations. However, diffi-
culties are engendered with this approach. For 
example, headwords are typically canonical in 
form whilst their corresponding lexical units in 
collocations can be variants (for example in-
flected or be derivations). Furthermore, in reality 
the definition of a collocation structure for lexi-
cographic purposes can itself be complex, for 
example to cater for or indicate information such 
as inflected forms, synonymy and translation 
equivalence, grammatical labelling and com-
ments (Gaveiro, 1998, pp. 26 - 27, 64 - 65).  
More recently, our interest has been concerned 
with how to develop such multilingual colloca-
tion dictionaries including access to collocations 
based on linguistic phenomena as well as by 
headwords (a headword can only be a single 
word, even for idioms) (Chan, 2005) where the 
issues of particular cases at the semantic level 
and at the grammatical level are important. Here 
32
the access to collocations can be by posing a 
problem; one can ask for those collocations 
which present a problem of article for example. 
The linguistic phenomena and lexicographic 
requirements are ones that are candidates for 
modelling such dictionaries using formal meth-
ods, for example using the Z formal specification 
language (Spivey, 1992), the impetus being that 
certain domains in which such dictionaries are 
used are safety critical in nature. This has re-
sulted in work in respect of the state invariants 
peculiar to specialised multilingual collocation 
based dictionaries (Greenfield, 1998a; 
Greenfield, 1998b). 
In response to these various observed linguis-
tic phenomena and lexicographical requirements, 
the MultiCoDiCT (Multilingual Collocation Dic-
tionary System Centre Tesnière) system was de-
veloped as a research aid tool for multilingual 
lexicographic research (Greenfield et al., 1999; 
Greenfield, 2003). The basic model underpinning 
MultiCoDiCT dictionaries reposes on the con-
cept of the collocation sense group as a means to 
ensure integrity and consistent access to the col-
locations. In this model a collocation in a 
language appears only once, whereas in 
conventional dictionary models it is the 
headword in a language that appears only once. 
This constraint leads us to generality; not only do 
we obtain reversibility of translation with no 
extra effort, we obtain non-reversibility of the 
lexis where this happens to be the case. Further-
more, headword access to a collocation also pro-
vides direct access to the other collocations in the 
dictionary with an equivalent sense (or senses for 
polysemic collocations). 
More recently, work on linguistic phenomena 
and lexicographic attributes based access to col-
locations (as well as headword access) has re-
sulted in a prototype system using an algorithmic 
approach (Chan, 2005) using the Studygram sys-
tem (Cardey and Greenfield, 1992). 
In the paper we first review the linguistic phe-
nomena and lexicographic requirements that we 
have discerned for such multilingual collocation 
dictionaries. We then discuss the development of 
such dictionaries in which the linguistic phenom-
ena and lexicographic attributes are themselves a 
means of access to the collocations. Finally, in 
the conclusion we show how Studygram and 
MultiCoDiCT can be integrated in order to pro-
vide a more general approach for the access to 
such multilingual collocation dictionaries. 
2 Linguistic phenomena and lexico-
graphic requirements 
In the context of lexicographical research, collo-
cations as articles in multilingual dictionaries 
present various linguistic phenomena and lexico-
graphic requirements which are sufficiently ge-
neric but also sufficiently important lexico-
graphically as to warrant some generalised sup-
port. The various phenomena and requirements 
are illustrated in this section by the essentially 
traditional headword access method to colloca-
tions as provided by the MultiCoDiCT system. 
The linguistic phenomena concern synonymy, 
polysemy and non-reversibility of the lexis in 
translation. For example synonymy is indicated 
by more than one collocation having the same 
sense equivalence variously in the source lan-
guage or in the target language (in the illustra-
tions that follow the source language is on the 
left and the target language is on the right); see 
Figures 1 and 2. 
 
Spanish French 
Headword  
boleto billet 
Collocations 
billete de avión (Spain) 
boleto de avión (Americanism) 
billet d'avion 
Figure 1. Synonymy in the source language 
 
French Spanish 
Headword  
billet billete(Spain) 
boleto(Americanism) 
Collocations 
billet d'avion billete de avión (Spain) 
boleto de avión (Americanism) 
Figure 2. Synonymy in the target language 
 
In the above two examples, the Spanish colloca-
tions include annotations indicating regionalisms 
such as (Spain) (Chan, 1999). We say that collo-
cations in the same or different languages which 
are equivalent in that they have the same sense 
are members of the same sense group. In the 
above examples we can also observe various 
lexicographical requirements such as headwords 
and the use of structured annotations to display 
the regionalism information.  
Polysemy is indicated by the presence of 
translations with different senses, that is, where a 
collocation is the member of more than one sense 
group. The example that we use is drawn from an 
archtypical bilingual dictionary (Werthe-
Hengsanga, 2001) of Thai-French image expres-
33
sions in the zoological domain with the particu-
larity that the types of translation are shown by 
lexicographic annotations as follows: 
− Eq equivalent – supplied, provided that an 
equivalent can be found 
− LT literal translation – word for word 
− AS analytical sense – literal translation re-
formulated in correct French 
− FS functional sense – the true sense of the 
translated collocation 
The Thai is shown here by means of a phonetic 
transcription using ASCII characters (which in 
fact does not provide an adequate cover but this 
matter is not pursued here). An example of a 
polysemic Thai collocation with 3 functional 
senses (FS) is shown in Figure 3. 
 
 Thai French 
Headword  
hmu: cochon, porc 
Collocations 
j?:n hmu: me: 
w: (TP) 
sense 1: 
donnant donnant (Eq) 
tendre porc<n(m,s)> tendre 
chat<n(m,s)> (LT) 
l'un tend son cochon<n(m,s)> 
l'autre son chat<n(m,s)> (AS) 
contre une chose, une prestation 
équivalente à ce qu'on donne 
soi-même (FS) 
sense 2: 
donnant donnant (Eq) 
tendre porc<n(m,s)> tendre 
chat<n(m,s)> (LT) 
l'un tend son cochon<n(m,s)> 
l'autre son chat<n(m,s)> (AS) 
prendre son dû séance tenante 
dans une transaction (FS) 
sense 3: 
donnant donnant (Eq) 
tendre porc<n(m,s)> tendre 
chat<n(m,s)> (LT) 
l'un tend son cochon<n(m,s)> 
l'autre son chat<n(m,s)> (AS) 
vendre et acheter comptant (FS) 
Figure 3. Polysemy illustrated by a Thai colloca-
tion with 3 functional senses (FS) 
 
The linguistic phenomenon ‘non reversibility 
of the lexis’ is illustrated by the example shown 
in Figure 4. 
 
French English 
Headword  
antécédents  'medical history' 
Collocations 
antécédents du patient patient history 
antécédents médicaux medical history 
 
English French 
Headword 
 
history 
− 
Collocations 
patient history antécédents du patient 
medical history antécédents médicaux 
Figure 4. Illustration of non reversibility of the 
lexis 
 
In this dictionary which is restricted to the do-
main of clinical research (Gavieiro 1998), even 
though there is a translation of the French head-
word antécédents by an English collocation 
'medical history' (printed between quotes to indi-
cate it to be a collocation rather than a headword 
in the target language), this is not the case for the 
inverse sense for the English headword history. 
Being a dictionary of collocations, the translation 
of history as a headword has no place in such a 
domain specific dictionary. On the contrary, 
English collocations containing the headword 
history have their place, they are translated to 
French. 
Lexicographic requirements can be divided 
into those which concern the functionality of-
fered by the dictionary (for example, as we have 
already seen, the use of annotations for various 
purposes) and those which concern the organisa-
tion and integrity of the dictionary. 
The functionality offered by such a dictionary 
includes the method of access to collocations as 
articles, the presentation of the articles in order 
to display any of the linguistic phenomena pre-
sent (as has been illustrated by the examples 
above concerning synonymy, polysemy and non-
reversibility of the lexis in translation), and the 
organisation and provision of lexicographical 
annotations. 
For the access to collocations as articles this 
can be as in conventional paper dictionaries by 
means of headwords, typically in alphabetic 
order. A headword is an individual lexical unit 
whose primary purpose is to provide a means of 
access to a collocation. In the MultiCoDiCT 
system a headword is never the whole 
collocation even for a fully fixed expression. A 
given headword can access several collocations 
(as is illustrated in Figure 4) and in like manner, 
34
a collocation can be accessed by many 
headwords. This can be seen for the collocation 
antécédents médicaux with headwords 
antécédents (Figure 4) and médical (Figure 5).  
 
French English 
Headword  
médical medical 
Collocations 
antécédents médicaux medical history 
Figure 5. Variant form of headword in the con-
text of the collocation 
 
The headwords of a collocation require to be 
specified; in the MultiCoDiCT system this is 
done explicitly by the lexicographer. Because of 
inflexional and derivational morphology, the 
headwords are typically in a canonical form, 
whilst the forms in the collocations can be vari-
ants; Figure 5 illustrates this for the French 
headword 'médical' which takes the variant form 
'médicaux' in the French collocation. In Figure 4, 
the case of the headword antécédents (nominally 
a ‘variant’ (plural) of the canonical form anté-
cédent) is atypical, the lexicographical choice of 
the form of the headword here being due to anté-
cédents being a short form of antécédents médi-
caux. Thus in the organisation of the dictionary 
there must be, as is the case in the MultiCoDiCT 
system, a mapping between headwords in their 
canonical form and their ‘presence’ in colloca-
tions. 
With annotations such as grammatical func-
tion (already shown in Figure 3) even the linguis-
tic phenomenon of grammatical variation can be 
accounted for, as shown in Figure 6. 
 
French Spanish 
Headword 
 
aérogare terminal<n(m,s)>  
'estación<n(f,s)> terminal 
<adj(f,s)>' 
Figure 6. Illustration of grammatical variation 
 
In the case of synonyms or polysemic equiva-
lences, a given word can ‘change’ its grammati-
cal role. In the first synonymic equivalence in the 
example in Figure 6, the Spanish word terminal 
is a noun whilst in the second it has as grammati-
cal function adjective because the word estación 
has as role a noun. It should be noted that for the 
two grammatical functions of the word form 
terminal, in the Spanish lexis there is only one 
headword for terminal. 
We now turn to the phenomena which have an 
impact on the organisation and integrity of such a 
dictionary and thus its underlying model and 
how this has been achieved in the MultiCoDiCT 
system. We must deal with variously colloca-
tions, headwords and annotations and the various 
interrelations between these such as sense groups 
and furthermore the relation between headwords 
and collocations, all these in a multilingual con-
text. There must necessarily be some means to 
ensure the integrity of these various data items 
and the relations between them. 
The model that underpins the MultiCoDiCT 
system is based on: 
• firstly the sense group to which one or 
more collocations across one or more lan-
guages is associated in being sense 
equivalent (a sense group is no more than 
such an association), 
• secondly the languages, to each of which 
collocations are uniquely associated, 
• thirdly the collocations and 
• fourthly the headwords, which in the 
MultiCoDiCT system are the only way to 
access directly a collocation and its sense 
equivalences (synonyms and translations). 
(Acces to collocations by means of lin-
guistic phenomena is discussed in the next 
section.)  
In respect of annotations, the underlying 
model allows these to be added at the level of 
sense group, collocation (for example regional-
ism) or collocation lexical item (for example 
grammatical category) 
 As far as the collocations which are the mem-
bers of a sense group are concerned, these can be 
viewed orthogonally over two dimensions. One 
dimension involves the language and here too 
one or more languages may have a special status, 
such as Latin in dictionaries of flora and fauna 
which we address in the next section of the pa-
per. The other dimension concerns the nature of 
collocations. Here we can type collocations as 
either being ‘true’ collocations in terms of the 
linguist's view, or, collocations which are trans-
lators’ paraphrases such as for example transla-
tion types as we have already discussed and 
shown (Figure 3).  
35
3 Linguistic phenomena and lexico-
graphic attributes as a means of ac-
cess 
Common Spanish 
names 
corozo agutí 
PANAMA corozo ñeque 
BOLIVIA totai - 
CUBA - jutía mocha 
MEXICO - cotuza
VENEZUELA corozo zuliano de
grupa negra 
French translation acrocome / 
coyol / noix 
de coyol 
agouti 
In this section we consider linguistic phenomena 
and by extension lexical attributes such as anno-
tations of linguistic phenomena as themselves a 
means of access to the collocations in multilin-
gual collocation dictionaries. We illustrate this 
approach by describing a bilingual dictionary of 
tourism that has been developed with this means 
of access in mind. 
Figure 7. The presence of varieties of languages 
 
We also found cases of linguistic phenomena 
at the semantic level, such as Americanisms, 
Anglicisms, non-reversible equivalents, etc. To 
handle these various observations we developed 
an algorithmic dictionary access method in order 
to provide access to the relevant collocations and 
translations. Our overall algorithm (see Figure 8) 
is itself composed of three principle sub-
dictionaries:  
This dictionary involves the differences be-
tween French-Spanish-French translations found 
in ecological, fluvial and cultural (historical and 
religious) tourism corpora (Mendez, 1993; Ri-
gole and Langlois, 2002). When translating the 
corpora, we noticed the presence of varieties of 
languages, such as Latin American Spanish and 
common Spanish (that is the Spanish of Spain) 
and of regionalisms; for example, in the case of 
Panama whose animal and plant specimen names 
were used only in that country and not in other 
Latin American countries, see Figure 7 (Chan 
2005). 
a. French-Spanish-French equivalences 
(537 words),  
b. particular cases at the semantic level 
(1146 words) and  
 
c. particular cases at the grammatical level 
(291 sub-conditions). 
 
 
 
 
 
 
 
 
 
 
 
Tourism dictionary overall algorithm
   
 
  
 
 
 
 
religious historical 
cultural tourism  
fluvial tourism 
ecological tourism 
equivalences 
(French-Spanish-French) 
synonymy
common and rhetorical senses
polysemy
paraphrases
proper nouns 
languagelevel
collocations
non-reversible equivalents 
Anglicisms
Americanisms
particular cases 
at the semantic level
gerund, past participle and  
verbal adjective  
gender and number 
demonstratives 
articles 
adverbs 
the place of adjectives 
particular cases 
at the grammatical level
 
 
 
 
 
grammatical category problems
 
 
verbal system 
relatives and indefinite pronouns
 
 
 
 
Figure 8.  Dictionary access algorithm 
 
36
The algorithm has a maximum of eight levels 
where the existence of other sub-dictionaries (or 
sub-algorithms) is possible inside of each dic-
tionary, which itself can be consulted independ-
ently or dependently. In other words, the overall 
algorithm includes several mini-algorithms and 
mini-dictionaries. 
At the start of each consultation, the words be-
longing to a given dictionary are presented in the 
form of a list arranged in alphabetical order so 
that the user can save time. 
We now discuss these three specific sub-
dictionaries.  
The first sub-dictionary concerns equivalences 
which are provided in the French-Spanish-
French languages and which are classified ac-
cording to topic. The sub-field cultural tourism 
presents for example historical and religious 
tourism as sub-sub-fields. 
The second sub-dictionary concerns particular 
cases at the semantic level, the terms of the dic-
tionary of the Panamanian fauna, for example, 
are joined together by class such as: insects, 
mammals, birds and reptiles. The user can check 
starting from: 
• French to obtain the equivalences in 
Spanish of Panama and common Spanish;  
• French to obtain the equivalences in 
common Spanish and Latin;  
• Panamanian Spanish to obtain the 
equivalences in common Spanish; 
• common Spanish to obtain the equiva-
lences in Panamanian Spanish;  
• Panamanian Spanish and common Span-
ish to obtain the equivalences in French 
and Latin;  
• Latin to obtain the equivalences in 
French, common Spanish and Panamanian 
Spanish. 
At the outset we had the intention to develop a 
bilingual dictionary. However, we included Latin 
in the dictionary, since, when translating the 
Spanish corpora to French, we noticed that the 
names of the flora and fauna belonged to a spe-
cialised lexicon and that most of these names 
constituted regional variations. Thus, we had to 
look for the scientific name (coming from Latin), 
then the common Spanish name in bibliographi-
cal documents, monolingual dictionaries or on 
Internet sites dedicated to these fields and finally, 
to look for the French translation in general bi-
lingual dictionaries (Spanish-French) and on 
zoological and botanical websites in order to 
validate the equivalences. 
We did not consider the variants of other 
Latin-American countries because in order to do 
so it would have been necessary to undertake an 
intensive research exercise into the matter and to 
have had the terms checked by specialists in the 
field studied. 
The third and last sub-dictionary deals with 
grammatical findings.  It is not only composed of 
words but grammatical rules and also examples 
in order to illustrate the different cases.  For this 
reason, we do not mention the quantity of words 
in the dictionary but rather the number of sub-
conditions in the algorithm. 
The algorithm that we have developed is inter-
actively interpretable by the Studygram system 
(Cardey and Greenfield, 1992) which also pro-
vides the user interface. To illustrate the trace of 
a user access using our prototype system with the 
dictionary access algorithm illustrated in Figure 
8, we take as entry the French collocation 'ama-
zone à front rouge' and where we are interested 
in the equivalences sub-dictionary and the par-
ticular cases sub-dictionary (see Figure 9). 
 
 
 
tourist 
context 
loro m. frente roja loro m. frente roja
equivalences 
sub-
dictionary 
 
 
 
 
 
 
 
 
 
 
 
 
Figure 9. Trace of a user access with as entry the French collocation 'amazone à front rouge' 
 
amazone  
à  
front rouge 
particular 
cases at the 
semantic 
level sub-
dictionary 
nonreversible 
equivalents 
collocations 
amazone f. : amazona f. (femme)
à prep. : a / de / en / para / por /que 
front m. : frente f. (partie du visage)
rouge adj. : rojo,ja adj., encar-
nado,da adj., colorado,da adj. 
37
4 Conclusion 
We have presented the essential linguistic phe-
nomena and the lexicographic requirements that 
we have discerned to be useful in the develop-
ment of multilingual collocation dictionaries. By 
‘collocation’ we mean not only set or fixed ex-
pressions including idioms, simple co-
occurrences of items and metaphorical uses, but 
also translators’ paraphrases. The basic model 
for such dictionaries as exemplified in the Mul-
tiCoDiCT system reposes on the concept of the 
collocation sense group as a means to ensure in-
tegrity and access. We have presented a novel 
access method to such dictionaries using the 
Studygram system which in essence provides 
access also based on much of the very linguistic 
phenomena and related lexicographical attributes 
that we have previously discerned and thus ena-
bling access to collocations by posing a problem.  
We conclude in showing how Studygram and 
MultiCoDiCT can be integrated in order to pro-
vide a more general approach for the access to 
such multilingual collocation dictionaries. In this 
approach, Studygram would provide the user 
interface and problem solving capability as de-
scribed in section 3 and MultiCoDiCT would act 
as a lexical resources server. 
At one level this approach would involve the 
essentially technical matter of standardising the 
mutual call mechanism (operational semantics) 
between Studygram and MultiCoDiCT. The 
Studygram system in any case supports algo-
rithm solutions (called operations) which can be 
procedure calls, which in this context would be 
to MultiCoDiCT. 
At another level this approach would involve 
formalising and standardising the linguistic and 
lexicographic terminology shared by the two sys-
tems. This level is thus concerned with including 
the lexicographical needs in the computational 
model. In respect of the semantics of the Multi-
CoDiCT component, the model underpinning 
MultiCoDiCT could be extended in a simple 
fashion to support explicitly the provision of lin-
guistic ‘headwords’ involving the intrinsically 
modelled linguistic phenomena of synonymy and 
translation equivalences, polysemy and non-
reversibility of the lexis. Access by conventional 
headwords is in any case already supported. The 
same MultiCoDiCT model provides annotation 
structures attached to the sense group, to the col-
location and to the collocation lexical item. 
However the semantics of the annotation content 
is the lexicographer’s and thus would involve an 
agreed semantics between the MuliCoDiCT and 
Stydygram components including the algorithm 
content concerning the machine interpretation of 
such annotation contents and lexicographic at-
tributes. 

References 
Cardey, S., Greenfield P. 1992. The `Studygram' 
Natural Language Morphology System: A First 
Step to an Intelligent Tutoring Platform for Natural 
Language Morphology in Proceedings of the UM-
IST ICALL workshop, 42-59, published by CTI 
Centre for Modern Languages, University of Hull, 
UK.
Cardey, S., Greenfield, P. 1999. Computerised set 
expression dictionaries : design and analysis, Sym-
posium on Contrastive Linguistics and Translation 
Studies (Université Catholique de Louvain, Lou-
vain-la-Neuve, Belgique) 5-6 February 1999. In 
Lexis in Contrast (B. Altenberg & S. Granger eds.), 
Benjamins. 
Chan, R., 2005, El diccionario de la flora y fauna pa-
nameña: propuesta de algoritmo para la solución de 
problemas de traducción de español-francés, IX 
Simposio Internacional de Comunicación Social, 
Santiago de Cuba, 24-28 de enero 2005, Actas I, 
389-393. 
Gaveiro, E., 1998, Elaboration d'un Prototype de Dic-
tionnaire Informatisé Français-Anglais / Anglais-
Français. Application au Domaine de la Recherche 
Clinique, Mémoire de D.E.A., Université de Fran-
che-Comté, France. 
Greenfield, P. 1998a. L'espace de l'état et les inva-
riants de l'état des dictionnaires terminologiques 
spécialisés de collocations multilingues, Actes de la 
1ère Rencontre Linguistique Méditerranéenne, Le 
Figement Lexical, Tunis, les 17-18 et 19 September 
1998, 271-283. 
Greenfield, P. 1998b. Invariants in multilingual ter-
minological dictionaries, BULAG N° 3, ISBN 2-
913322-11-5, Presses Universitaires Franc-
Comtoises, 1998, 111-121. 
Greenfield, P. 2003. Le rôle de l'informatique dans le 
traitement et l'enseignement des langues, Actes du 
Congrès international : Journées linguistiques 
franco-asiatiques, Naresuan University, Phitsanu-
lok, Thaïlande, 20-22 August 2003, 69-84. 
Greenfield, P., Cardey, S., Achèche, S., Chan Ng, R., 
Galliot, J., Gavieiro, E., Morgadinho, H., Petit, E. 
1999. Conception de systèmes de dictionnaires de 
collocations multilingues, le projet MultiCoDiCT, 
Actes du Colloque international VIème Journées 
scientifiques du Réseau thématique de l'AUF Lexi-
cologie, Terminologie, Traduction, Beyrouth, 11-
13 November 1999, 103-113. 
Mendez, E. 1993. Los roedores de Panamá.  Panamá, 
Laboratorio Conmemorativo Gorgas, pp.59-64 and 
pp.281-286. 
Moon, R. 1998. Fixed expressions and idioms in Eng-
lish, a corpus-based approach. Clarendon Press, 
Oxford. ISBN 0-19823614-X. 
Rigole, M., Langlois, C-V. 2002. Panamá.  Guides de 
voyage ULYSSE. 4th edition.  Canada, Guides de 
voyage Ulysse inc., 333p. 
Spivey, J.M. 1992. The Z Notation. Prentice Hall, 
ISBN 0-13-978529-9. 
Werthe-Hengsanga, V. 2001. Etude de la traduction 
automatique en français des expressions imagées 
de la langue thaï (domaine animalier). DEA, 
Sciences du langage, Université de Franche-Comté, 
Besançon, France. 
