Combining the semantics of collocations with situation

4 downloads 0 Views 747KB Size Report
situation-driven search paths in specialized dictionaries. Marie-Claude L'Homme and Patrick Leroyer. The systematic presentation of collocations is increasingly ...
John Benjamins Publishing Company

This is a contribution from Terminology 15:2 © 2009. John Benjamins Publishing Company This electronic file may not be altered in any way. The author(s) of this article is/are permitted to use this PDF file to generate printed copies to be used by way of offprints, for their personal use only. Permission is granted by the publishers to post this file on a closed server which is accessible to members (students and staff) only of the author’s/s’ institute, it is not permitted to post this PDF on the open internet. For any other use of this material prior written permission should be obtained from the publishers or through the Copyright Clearance Center (for USA: www.copyright.com). Please contact [email protected] or consult our website: www.benjamins.com Tables of Contents, abstracts and guidelines are available at www.benjamins.com

Combining the semantics of collocations with situation-driven search paths in specialized dictionaries Marie-Claude L’Homme and Patrick Leroyer

The systematic presentation of collocations is increasingly recognized as a very useful addition to specialized reference works. However, few dictionaries or terminological databases actually include this kind of data. More surprisingly still, no method has been designed yet to allow efficient access to and retrieval of specific specialized collocations from electronic reference tools. This article presents two new search paths for accessing and extracting collocations from an English-French specialized lexical database. The paths have been designed according to two specific user-defined situations: (1) translation from L1 to L2; and (2) text production in L2. We exploit a formal semantic encoding of collocations based on Lexical Functions (LFs). LFs allow us to establish an equivalence relationship between collocations that convey the same meaning in different languages without having to link the collocations formally. They also allow us to extract sets of collocations associated with specific meanings. Keywords: collocation, terminology, specialized lexicography, equivalent, lexical relation, lexical function, search paths, lexicographic functions

1. Introduction The systematic presentation of collocations is often recognized as an extremely useful addition to specialized dictionaries (Bergenholtz and Tarp 1995; Heid and Freibott 1991, Schneider 1998, among others). The listing of collocations is viewed as a solution to the gap between needs raised by the production of texts (especially in L2) as well as in L1 to L21 translation — two communicative activities in which terms need to be placed in a linguistic and discursive environment (Leroyer 2001) — and the actual information provided by specialized dictionaries (often reduced to short articles specifying the conceptual content of nouns and noun phrases). Terminology 15:2 (2009), 258–283.  doi 10.1075/term.15.2.05lho issn 0929–9971 / e-issn 1569–9994 © John Benjamins Publishing Company



Combining the semantics of collocations with situation-driven search paths 259

Although this is recognized, there does not seem to be a general agreement as to what types of word combinations should be listed, nor as to how they should be presented in specialized reference works. Different forms of specialized word combinations are taken into consideration by authors with more or less formal definitions. They range from short or long phrases restricted to those sequences that will likely be useful in specific user-defined situations (Nielsen and Mourier 2007) to predicate-argument relationships with restrictions on the arguments (Lerat 2008; Pedersen 1998), including binary combinations whose status as collocations is defined according to criteria provided by linguistics or lexical semantics (Cohen 1986; Heid and Freibott 1991, among others). The presentation of word combinations in specialized reference works is also diversified as will be seen in Section 3. More surprisingly still, even at a time when an increasing number of resources in electronic form are available, no efficient method has been designed yet to retrieve collocations from the entries (other than entering the components of the collocations which would simply allow users to know they exist). This article does not address the first problem, leaving it to terminologists or specialized lexicographers to define the word combinations they find relevant with regards to the needs their reference works are designed to fulfill.2 Rather, we focus on the second problem, namely that of designing efficient methods to help users locate specific word combinations in specialized reference works. Our focus is on tools in electronic form (computerized dictionaries or terminological databases) which are viewed here (along with Sager 1990) as the only media that allow efficient organization of and access to lexical data. More specifically, we designed two search paths for extracting collocations from an existing database in which French and English collocations are listed and formally encoded according to their semantics. This encoding is exploited in order to retrieve specific collocations in two pre-defined situations: 1. Translation of a specific collocation from L1 to L2, e.g. what is the French equivalent for hit a key? 2. Direct production in L2 of a collocation in which the collocate conveys a specific meaning, e.g. how is the meaning “start using” expressed with the terms file, computer, or program? The article is structured as follows. Section 2 gives a brief description of specialized dictionaries and databases that list collocations and some of the limitations users may encounter when accessing them. Given the focus of our work, a special emphasis has been placed on terminological repositories in electronic form. Section 3 presents our arguments for a retrieval process taking into consideration specific needs that can be fulfilled through a rich encoding of collocations. Section 4 provides some background on the linguistic framework on which our method is

© 2009. John Benjamins Publishing Company All rights reserved

260 Marie-Claude L’Homme and Patrick Leroyer

based, i.e. Lexical Functions (LFs). Section 5 describes the two search paths we designed. A few concluding remarks are made in Section 6, in which we also suggest directions for future work. 2. Specialized dictionaries and databases with collocations Even though most terminologists, specialized lexicographers and users of their products agree that collocations in dictionary articles are useful, few specialized resources actually contain word combinations. In addition, as was mentioned above, there does not appear to be a consensus as to the types of combinations to be listed or how they should be organized. Below is a short list of dictionaries that include word combinations and a brief description of how these are presented.3 – Cohen (1986): Specialized dictionary listing collocations in the field of stock exchange; French collocates are organized in tables according to their part of speech (noun, verb, adjective) and their general meaning (“increase”, “reduction”, etc.). – Meynard (2000): Specialized dictionary listing word combinations in the field of the Internet; collocates in English and in French are organized according to their part of speech with immediate access to the translations of collocations in each language (the dictionary is divided into two parts in which head words are ordered alphabetically in French, first part, or in English, second part). – Termium (2009): Term bank containing approximately 3 millions terms related to all sorts of special subject fields; collocates are listed in a limited number of term records and are organized according to their part of speech. – DAFA (2000), printed version:4 A dictionary for learners of the business language; French collocates are organized according to their part of speech, their argument structure, and their general meaning. In printed dictionaries (and in Termium as well), access to collocations is only allowed through the headword. Electronic dictionaries and databases can of course offer other ways of accessing collocations and, hopefully, more flexible ones. However, we will see that this is not always the case. In the following subsections, we describe how collocations are listed in three electronic specialized resources, what kind of data is provided on them and how this information can be accessed. Of course, these resources contain other data categories and functionalities, but these will not be listed or discussed here.5

© 2009. John Benjamins Publishing Company All rights reserved

Combining the semantics of collocations with situation-driven search paths 261



2.1 Dictionary of Accounting (DA) The Dictionary of Accounting, henceforth DA (Nielsen 2006; Nielsen and Mourier 2007), contains Danish and English terms as well as lexical units pertaining to the field of accounting. Along with other data categories, articles contain a section for collocations. Figure 1 is the reproduction of an article extracted from the English -> Danish DA. estimate 1 Noun estimat Noun Definition An estimate is the result of an approximate calculation. Collocations – estimate of future cash flows estimering af fremtidige pengestrømme – a reasonable estimate et rimeligt estimat – a changed estimate et ændret skøn – make a formal estimate of the recoverable amount foretage et formelt skøn over genindvindingsværdien – a reliable estimate et pålideligt skøn – highly subjective estimates meget subjektive skøn – explicit and unbiased estimates eksplicitte og neutrale skøn Examples – The amortisation period is the best estimate of the asset’s useful life. Afskrivningsperioden udgør det bedste skøn over aktivets brugstid. Synonyms skøn

Figure 1.  Article estimate (noun) in the English->Danish DA

In the DA, collocations are defined as typical phrases containing the headword that are selected according to user needs (Nielsen and Mourier 2007). They correspond to short or longer prototypical phrases that are frequently found in specialized texts associated with a specific genre or that might raise some translation problems.6 In addition, many headwords are polylexical units that can be compositional or semi-compositional. This means that items appearing as headwords

© 2009. John Benjamins Publishing Company All rights reserved

262 Marie-Claude L’Homme and Patrick Leroyer

in the DA can also be listed as collocations (e.g., long-term liability appears as a headword in an article, but is also listed as a collocation under liability). In addition, the same collocation (or similar ones) can be found in different articles. For example, reliable estimate appears under estimate; a reliable estimate of obligations is listed under reliable. In some articles, lists of collocations are longer than the one given for estimate (cf. Figure 1); articles contain between 20 and 30 such phrases (e.g., account, asset, bank). The article for tax lists up to 45 phrases. However, when browsing through articles, it is difficult to understand on what basis collocations are organized or sorted. If lists are long, users might experience problems accessing the specific combination for which they are looking. As can be seen in Figure 1, all collocations are given in the source language (English in this version of the DA) and in the target language (a reasonable estimate -> et rimeligt estimat). This is particularly useful for users looking for ways to handle phrases in which a term can be found in the specialized field of accounting and produce an equivalent and idiomatic expression in the target language. As far as access is concerned, for the time being, collocations cannot be extracted directly from the dictionary. They can be consulted by users once they have extracted an article for a given head word. In other words, reliable estimate can only be displayed once the articles for reliable or estimate are extracted from the database. 2.2 DAFA (2009), electronic version The online version of the Dictionnaire d’apprentissage du français des affaires, henceforth DAFA (DAFA 2009), contains terms related to the field of business. Although it includes terms in French, Dutch, English, Spanish, Italian, and German, collocations are given for French terms only. In the DAFA, collocations are not displayed all at once. Users must select a specific meaning (if the term is polysemic) and collocations are listed under different meanings, a part of speech category, and finally a labeled subdivision (if the collocates are adjectives or nouns, or a table, if the collocates are verbs). The name of the subdivision is hyperlinked: by selecting it, users visualize the corresponding list of collocations. Figure 2 shows part of the collocations given for contrat (contract) (when it refers to a written agreement) once the user choses noun collocates under the “Type” subdivision. For each collocation, an example and/or explanation is given. Verbal collocates are displayed in a table according to a classification based on lexical functions (cf. Section 4.1) that takes into account their meaning and the participant involved (Figure 3).

© 2009. John Benjamins Publishing Company All rights reserved



Combining the semantics of collocations with situation-driven search paths 263

Contrat + adjectif Sens 1.1. : Convention par laquelle une personne s’engage à (ne pas) faire qqch. => caractérisation, type Sens 1.2. : Acte qui enregistre cette convention. => type + nom Sens 1.1. : Convention par laquelle une personne s’engage à (ne pas) faire qqch. => mesure, type, autre Sens 1.2. : Acte qui enregistre cette convention. => type + verbe Sens 1.1. : Convention par laquelle une personne s’engage à (ne pas) faire qqch. => Tableau 1 Sens 1.2. : Acte qui enregistre cette convention. => Tableau 2

COLLOCATIONS contrat (sens 1.1.) : Convention écrite ou orale par laquelle une ou plusieurs personnes physiques ou morales (X) s’engagent envers une ou plusieurs autres personnes physiques ou morales (Y) à transférer une propriété, à faire ou ne pas faire qqch. (p. ex. couvrir un risque). contrat + nom, fonction type : Un contrat de + nom qui indique l’objet de l’accord. Un contrat de franchise ; de soustraitance. Un contrat de travail : contrat par lequel le salarié s’engage à mettre à certaines conditions (durée, …) son activité à la disposition de l’employeur en contrepartie d’une rémunération. (Syn. : un contrat de louage de services). (Ant. : un contrat de louage d’industrie de louage d’ouvrage). (V. location, 1). Toute l’opération de rupture d’un contrat de travail s’entoure d’un formalisme dont le but est de protéger les droits des deux parties, mais avant tout de la partie par définition la plus faible, à savoir le travailleur. Un contrat de travail à durée déterminée (un CDD). >< Un contrat de travail à durée indéterminée (un CDI). (Syn. : (moins fréq.) un contrat d’emploi). Un contrat de licence : contrat par lequel le titulaire d’un droit de propriété industrielle (un brevet, une marque, …) accorde à un tiers, en tout ou en partie, la jouissance de son droit d’exploitation et ceci à certaines conditions. Un contrat d’exclusivité : contrat signé avec un seul ou un nombre limité de parties. (Syn. : (moins fréq.) un contrat exclusif ). Le constructeur automobile allemand a signé un contrat d’exclusivité avec plusieurs clubs de football prestigieux en Europe.

Figure 2.  Noun collocations listed under contrat in the DAFA

© 2009. John Benjamins Publishing Company All rights reserved

264 Marie-Claude L’Homme and Patrick Leroyer COLLOCATIONS contrat (sens 1.1.) : Convention écrite ou orale par laquelle une ou plusieurs personnes physiques ou morales (X) s’engagent envers une ou plusieurs autres personnes physiques ou morales (Y) à transférer une propriété, à faire ou ne pas faire qqch. (p. ex. couvrir un risque). contrat + verbe : X et Y

la négociation d’un contrat

Signer un contrat conclure un contrat passer un contrat avec X (Y) décrocher un contrat obtenir u  n contrat remporter un contrat Placer un contrat auprès de Y souscrire un contrat d’assurance (auprès de X)

la signature d’un contrat la conclusion d’un contrat – – l’obtention d’un contrat – – la souscription d’un contrat d’assurance (auprès de X)



négocier un contrat

X et Y Y (X) X

X Y



1 2 3

4

Figure 3.  Verbal collocates listed under contrat in the DAFA

In its current state, although the DAFA contains data of a high quality, users may experience problems accessing it due to the linguistically determined data presentation choices that have been made by the authors. Collocational data are not accessible directly but have to be selected from menus featuring semantic discrimination of entries together with an analytic classification of collocational data in accordance with the syntactic patterns of the base and its collocates (+noun, +adjective, +verb, etc.) and in accordance with a very general semantic classification (type, characterization, measure, level, others, etc.). Although simplified, the lexical classification is in line with the linguistic description developed by Mel’čuk (1984–1999) in the framework of the Meaning-Text-Theory (more details on this framework are given in Section 4.1). As already examined in detail in Leroyer (2006), priority in the DAFA seems to be given to knowledge acquisition of the features of the specialized language of business by way of fine-grained representation of the collocational data. On the other hand, practical solutions to communicative problems are underestimated, especially text production in L2, in the foreign language. For the non-native user, the general problem remains the easy access to the information needed. The abstract quality of the onomasiological arrangement, the interlinked semantic networks and the syntactic and semantic valency and arguments patterns also imply that users should be quite familiar with this linguistic presentation and its terminology. In brief, the fragmentation of access to collocational data and the absence

© 2009. John Benjamins Publishing Company All rights reserved



Combining the semantics of collocations with situation-driven search paths 265

of semasiological ordering jeopardize fast and easy access to the collocational data types needed for text production by a much larger user group. 2.3 DiCoInfo The DiCoInfo (Dictionnaire fondamental de l’informatique et de l’Internet) (L’Homme 2009a, 2009b) contains terms, mostly French, pertaining to the fields of computing and the Internet. An increasing number of English entries are in the process of being added. Along with other data categories, articles contain a section for collocations. Figure 4 shows an article extracted from the English DiCoInfo.7 In this resource, collocations listed are binary combinations containing the headword in which the components share a syntactic and semantic relationship (Hausmann 1979; Mel’čuk et al. 1995). Collocations (defined as syntagmatic relations) are listed in a table along with paradigmatic relations (file in the table apattachment1, n

Status: 2

Actantial structure: an attachment: ~ with Patient{file 1} sent by Agent{sender 1} to Destination{recipient1} in Support{email 1} Linguistic realizations of actants8 Synonym(s): attached document, attachment document, attached file, attachment file, email attachment Contexts Attachments: Your e-mail client allows you to add attachments to e-mail messages you send, and also lets you save attachments from messages that you receive. (Source : EMAIL) / If you then Indulge in risky practices such as downloading files from the Internet or opening e-mail attachments, infection is a matter of when, not if. (Source : CALL THE PC DOCTOR) / To send an attachment, compose the email and then, before sending it, hit the Attachment button in your mail program. (Source : PC101) Lexical relations Generic That contains a virus The sender puts an a. in the email The sender removes an a. from the email The sender sends the file to the recipient as an a. The recipient starts using an a. The recipient uses an a. French: pièce jointe 1

Figure 4.  Article attachment extracted from the DiCoInfo

© 2009. John Benjamins Publishing Company All rights reserved

file 1 infected 1 ~ add an ~ to … delete 1 an ~ send … as an ~ open 1 an ~ receive an ~

266 Marie-Claude L’Homme and Patrick Leroyer

pearing in Figure 4 is defined as a hyperonym).9 In addition to being listed, lexical relations are explained by means of a short paraphrase. A hyperlink to the entries describing the related term or the collocate (e.g., infected, file, delete) is provided when the corresponding entry is online. The DiCoInfo provides two basic access paths (below: Search modes) to the contents of its entries. First, the selection of Term will find the character strings (complete or partial) in the headword or synonym section of the entry. Enter term: delete



Search mode: • Term • Lexical relation Language: • French • English • Bilingual Precision: • Exact term • Starting with • Containing

1. Search mode : Term delete1, vt Actantial structure : delete: Agent{user 1} ~ s Patient{character 1, file 1} from Source{storage device, text}

Status: 2

Linguistic realizations of actants Contexts Windows are used to: share common attributes through out the entire system contain menus for File, Edit, View and Help which change depending on the task File open, rename, delete or change properties of files and folders (Source : INTROCOMP) / If you receive an email from someone you don’t know that has an attached file DO NOT OPEN THE FILE, delete it immediately. (Source: COMPEBEGIN) / Delete the currently selected email(s) from the currently selected folder. (Source INTERNET) Lexical relations Near synonym Antonym Noun

Remove create 1 deletion 1

French: supprimer1 2. Search mode : Lexical relation attachment 1: delete 1 an ~ (The sender removes an a. from the email) create 1: delete 1 (Antonym) deletion 1: delete 1 (Verb) email 1: delete 1 an ~ (The recipient destroys an e.) file 1: delete 1 a ~ (The user destroys a f.) folder 1: delete 1 a ~ (The user destroys a f.) icon 1: delete 1 an ~ (The user destroys an i.) partition 1: delete 1 a ~ (The user removes a p.) type 1: delete 1 (Opposite)

Figure 5.  Result of queries on delete using Term search mode and Lexical relation search mode

© 2009. John Benjamins Publishing Company All rights reserved



Combining the semantics of collocations with situation-driven search paths 267

Secondly, the selection of Lexical relation will look for character strings listed in the lexical relations section of the entry. Figure 5 shows the results of two queries with delete using the first and the second option. In the DiCoInfo, collocations are listed in English and French entries but equivalent collocations, even if they appear in entries, are not formally related (as is the case in the DA (Section 2.1)). Hence, someone looking for the translation of send … as an attachment will need to access pièce jointe (given as the equivalent of attachment), go through all the collocations listed in this entry, and decide that envoyer … en pièce jointe or transmettre … en pièce jointe are relevant translations. Finally, collocations are organized in the DiCoInfo according to a model proposed in L’Homme (2009a) in which collocates are ordered in an entry according to distinctions based on lexical functions (Mel’čuk et al. 1995) and the typical sequence of use of the object denoted by the base of the collocations. This will be further described in Section 4.1. However, even though collocations are organized according to such principles, lists in certain entries can be very long (in the French entry for fichier (file), nearly 150 lexical relations are listed, among which most are collocates) and thus users might experience frustrations when searching for a specific collocation. 2.4 Access problems related to specific user situations Up to now, we have shown that, although criteria for selecting relevant specialized word combinations and methods for listing them have been applied in dictionaries and databases (even if these criteria and methods vary), currently, there is no strategy for extracting sets of collations in response to specific queries. For example, in the three specialized resources reviewed in Sections 2.1 to 2.3, there is no way of obtaining straightforward results for the two following queries: 1. The translation of a specific collocation: In the DA, translations of phrases are provided, but users must first access the entry and go through a list before they locate the phrase or the portion of the phrase that they are looking for. In the DiCoInfo, users would need to go through the lists of collocations given for a term in L1 and its equivalent in L2 until they locate the relevant piece of information. The DAFA provides collocations only for French terms. 2. Search for collocates used to express a given meaning when combined with a specific term: In the DAFA, users can load web pages in which specific sets of collocates are listed (e.g., adjective collocates that express “a type of ” + term). In addition, verbs are classified according to their meaning (“increase”, “decrease”) and the participant involved in the collocation. But users must read the entire list to locate a collocate that expresses a given meaning. The same

© 2009. John Benjamins Publishing Company All rights reserved

268 Marie-Claude L’Homme and Patrick Leroyer

applies to the DiCoInfo: short explanations are given for each collocation, but users must read long lists before they locate the combination that is relevant for them. In the DA, users could not find answers to this sort of query. In other words, in both scenarios, users would need to go through more than one step to obtain the information for which they are looking. The burden on users can be reduced by limiting the number of phrases to be listed in entries. However, this presupposes that lexicographers or terminologists compiling the entries are capable of identifying beforehand the word combinations that are likely to fulfill the needs of users. While this can be done in resources where situations and user needs were defined in a very specific way (in the DA, for, example), it is extremely difficult to imagine such a strategy in a resource that aims to describe the vocabulary in less well defined fields (as is the case in the DAFA or in the DiCoInfo). In these latter cases, it would be very difficult, if not impossible, to anticipate the kinds of collocations that are likely to be searched in the future. The drawback of these resources is, of course, having long lists of word combinations in entries. This requires that some form or organization be defined,10 but this organization might not be transparent for users. 3. Access solutions based on user situations Our aim is to provide solutions to the limitations discussed in Section 2.4 and provide users with search paths that enable them to retrieve specific collocations according to pre-defined needs. Therefore, paths have to be implemented according to a functional framework, in which the point of departure is the actual user situation that triggers the need for consulting a dictionary. 3.1 User situations and the functional framework of specialized dictionaries In the functional framework of lexicography, dictionaries (and all other types of lexicographically designed sources of reference for that matter) are regarded and treated as utility products. Basically they are information tools. In precisely the same way as for all other kinds of tools, they are designed to meet specific information needs, and are therefore solely defined according to the function(s) they are meant to fulfill. The concept of “lexicographic function” is based on the causal relation between an extra-lexicographic user situation and an intra-lexicographic user situation, including the following elements: the specific user, user profile and specific user needs, the specific user situation, and the access to the specific data types from which the user will be able to extract needed information. According

© 2009. John Benjamins Publishing Company All rights reserved

Combining the semantics of collocations with situation-driven search paths 269



to Tarp (2008) a clear distinction must be made between two types of user situations: – The extra-lexicographic user situation which is a potential user situation expressed by the identification of lexicographically relevant user needs leading to consultation. – The intra-lexicographic user situation which is the real use situation of the dictionary and is expressed by the consultation of primary and secondary data types. The concept of “dictionary assistance” establishes the link between extra- and intra-lexicographic user situations, assistance being provided by gaining access to the needed data types from which information subsequently can be extracted. Primary data types are those containing the needed information for assistance, whereas secondary data types are those that are being used to guide the user towards the primary data types, and also help the user interpreting them. In the case of the dictionaries and databases examined in Section 2, collocations belong to the category of primary data types whereas headword entries and associated lexical relations should be regarded as guiding as well as interpretive elements belonging to the category of secondary data types. Furthermore, lexicography in a functional framework makes a necessary distinction and separation between cognitive functions on the one side, and communicative functions on the other. a. Cognitive functions and user situations are determined and motivated by the need to acquire new knowledge or check on existing knowledge. Dictionary consultation is normally of a sporadic kind but can also be of a more systematic kind, particularly in learners’ situation, in connection with the learning process of a foreign language or a specialized language for instance. Dictionary consultation, as explained in Bergenholtz (2005), can best be illustrated by the following consultation scheme:

user → dictionary → user

In cognitive user situations, lexicographically relevant needs are thus needs for encyclopedic knowledge related to language, specialized language, culture, or any specific subject field.

b. Communicative functions and user situations are determined and motivated by the need to get dictionary assistance when the user is being engaged in some textual activity, such as reading or revising a text, translating a source text into a target language, or writing a text in the mother tongue or in a foreign language. The consultation process is triggered by a specific text segment

© 2009. John Benjamins Publishing Company All rights reserved

270 Marie-Claude L’Homme and Patrick Leroyer

revealing specific problems to the user. The consultation scheme is radically different:

user → text segmentx → user → dictionary → user → text segmentx

3.2 New access paths The two needs we identified for the design of our new access paths — translation from L1 to L2 and text production in L2 — belong to the category of communicative user situations and functions described in 3.1 above. Our access paths should allow users to find straightforward answers to the type of query illustrated by the two following examples: 1. Search for the translation of a specific collocation: A user wants to access the French translation of send something as an attachment (envoyer … en pièce jointe; transmettre … en pièce jointe). 2. Search for collocates used to express a given meaning when combined with a specific term: which are the collocates that express the idea of “use” when combined with file (edit a ~), with mouse (move a ~; click on … with a ~, drag and drop … with a ~), or with Internet (browse the ~, search … on the ~). The search paths, as we designed them, rely on two prerequisites: the semantics of the collocations must be described in the database in a very systematic way, and must be accessible to the search engine providing access to collocations. Hence, they are applied to the DiCoInfo database since it complies with two of their requirements: it contains collocations in more than one language, and it provides information on the semantics of collocations. In addition, we will assume no prior linguistic knowledge from the user. Our access paths are designed so as not to bother users with questions related to the linguistic properties of terms or collocates (i.e., part of speech of collocates; syntactic position of terms when combined with collocates). As was mentioned above, specialized lexicographers and terminologists have different views on what constitutes a relevant word combination to be listed in a dictionary entry. However, users of specialized resources are likely to look for different types of sequences, and not necessarily binary combinations of words. The DiCoInfo lists binary combination of terms; hence, for practical reasons, access is limited to these types of sequences.11

© 2009. John Benjamins Publishing Company All rights reserved



Combining the semantics of collocations with situation-driven search paths 271

4. Exploiting the semantics of collocates The two paths we designed are enrichments of a model proposed in L’Homme (2009a) in which collocates are ordered in an entry according to distinctions based on lexical functions (Mel’čuk et al. 1995) and the typical sequence of use of the object denoted by the base of collocations. This model has been defined as a method for encoding and listing collocations in entries (in fact, as a series of criteria to be used by lexicographers when compiling entries). We will now convert it in order to exploit its explanatory power for the purpose of extracting specific collocations. In other words, we will adapt it so that it meets specific user needs and situations. Before we look at the concrete search paths we designed, we will give some background on the basic principles on which they are based: namely, Lexical functions (LFs), a formal system for encoding collocations designed within Explanatory and Combinatorial Lexicology (ECL). 4.1 Lexical functions and specialized lexicography Lexical functions are parts of a system designed to represent a large set of lexical relations. They have been proposed by Explanatory Combinatorial Lexicology, ECL (Mel’čuk et al., 1995; Mel’čuk 1984–1999), a theory of the lexicon which is part of a broader linguistic theory called Meaning-Text Theory. It aims at providing a systematic and formal description of the entire set of linguistic properties of lexical units, defined as lexemes or phrasemes that convey a specific meaning. A lexical function (LF) is written f(x) = y: f represents the function, x, the argument, and y, the value expressed by the function when applied to a given argument. The meaning of an LF is abstract and general and can be associated with a relatively high number of values. For example, Magn is a function that expresses intensification. It can be applied to different lexical units and produce a high set of values (e.g. Magn(smoker) = heavy; Magn(bachelor) = confirmed, etc.) (Mel’čuk et al. 1995: 126–127). There is a set of approximately 60 standard lexical functions divided into paradigmatic and syntagmatic functions (Mel’čuk 1998; Mel’čuk et al. 1995; Wanner 1996). LFs can also be combined to account for more complex meanings. As far as collocations are concerned, lexical functions encode three different types of linguistic properties: 1. The syntactic relationship between the base and the collocate. For example, the verbal collocates of space bar, namely press, release and insert, are distinguished based on the fact that, when used in combination with press or release, space bar is the first complement (direct object); when used in combination with insert, space bar is the second complement.

© 2009. John Benjamins Publishing Company All rights reserved

272 Marie-Claude L’Homme and Patrick Leroyer



press the ~: verb + 1st complement release the ~: verb + 1st complement insert … (a space) with the ~: verb + 2nd complement

2. The actantial structure of the base. Many collocations, especially verbal and nominal collocations, convey a meaning that involves one or more actants that appear in the actantial structure of the base. Another example of press and insert being used in combination with space bar is given below (Figure 6). The sense of press the space bar involves the first actant of space bar; the sense of insert … with a space bar, involves both actants of the term. space bar, n Someone (1st actant) presses the ~

Actantial structure: ~ used by someone (1st actant) to act on something (2nd actant)

Someone (1st actant) inserts something (2nd actant) with the ~

Figure 6.  Relationships between actantial structure and collocations

3. The meaning of the collocate. Collocates convey different meanings when combined with specific terms. For example, press, release and insert refer to typical uses of the space bar; however, other collocates convey a meaning of “creation”, e.g., create a file, a password, develop, write a program. Using LFs, press the space bar, release the space bar, and insert something with a space are encoded as follows:

Real1 (space bar) = press the ~ FinReal1 (space bar) = release the ~ Labreal12 (space bar) = insert … with the ~

Real is used when the base is first complement while Labreal is used when the base is second complement. Numbers in subscript correspond to the actants involved in the collocation. Real1 means that the first actant of space bar is the subject of press. The same applies to FinReal1. Finally, in the context of space bar, Real and Labreal express “use a space bar”. Fin is a processual function that is generally combined with another LF to express “end”.

© 2009. John Benjamins Publishing Company All rights reserved



Combining the semantics of collocations with situation-driven search paths 273

In the DiCoInfo, lexical functions have been used for two different purposes: 1. as a means to classify collocations according to their meaning and their structure (actantial chiefly); and 2. to order them. The classification according to the meaning and structure of collocates will lead to having different collocations appear in the same part of the list (e.g., in the entry for password, good ~ and valid ~ will appear next to each other; the same applies to enter a ~ and type a ~). The ordering is done based on the following criteria: 1. Collocations that refer to a “Type of ” something;

Password: bad, good, valid ~

2. Verbal collocates are ordered according to the following principles: a. The participant involved (e.g., in the entry for password, collocations associated with the owner of the password — created a ~, lose a ~ — will be listed together; those associated with the entity requesting a password — ask for a ~, accept a ~ — will appear somewhere else). b. The sequence of use of the thing denoted by the base of the collocation (e.g., collocations associated with the owner of a password will appear thus: create a ~, obtain a ~, have a ~, enter a ~, access … with a ~, etc.). Although this method is currently being implemented throughout the DiCoInfo and proves useful for terminologists to systematize their encoding, it remains difficult for users to make sense of the long lists of collocations provided in some entries. However, as it will be seen in the following section, this encoding lends itself to more user-specific needs with regards to accessing relevant collocations. 5. Implementing two new search paths The remainder of the article presents the two paths we devised to access collocations in electronic specialized resources. As was previously mentioned, access is tested and illustrated on the DiCoInfo database. However, it could be extended to other resources, provided that they describe more than one language, and that the semantics of collocations is described in a systematic way.12 5.1 Search path 1: translation L1 to L2 An access to equivalent collocations in different languages can be permitted even though collocations are not linked formally, provided that the system used to describe them is the same in all languages and that it discriminates collocations according to their linguistic properties. This is possible using the lexical functions system.

© 2009. John Benjamins Publishing Company All rights reserved

274 Marie-Claude L’Homme and Patrick Leroyer Enter term: attachment Combined with: send Results Send … as an attachment -> envoyer … en pièce jointe, transmettre … en pièce jointe

Figure 7.  Search path for the L1-L2 translation function

In the DiCoInfo, users could enter a collocation in a specific language, e.g., English, and find the equivalent in a different language, e.g., French. Figure 7 shows how a query combining attachment and the verb send produces collocations in the two languages. In order to produce these results, the system first needs to search for attachment and its equivalent (pièce jointe).13 Then, to establish the equivalence relationship between collocations, the system must access the formal encoding of collocations. The results reproduced above are obtained, since the collocation in English and its equivalent in French are encoded with the same lexical function, namely Labreal123 (the access to LFs must be carried out in a transparent manner so that users would not see this formal encoding). The process is illustrated below. Enter term: attachment Combined with: send Step 1: Find the term entered by the user: attachment Step 2: Locate the entry with the equivalent term: pièce jointe14 Step 3: Locate the specific lexical unit entered by the user: send Step 4: Extract the lexical function that encode the relationship between send and attachment: Labeal12315 Step 5: Locate the same lexical function in the record for pièce jointe.16

Results Send … as an attachment -> envoyer … en pièce jointe, transmettre … en pièce jointe

The type of query illustrated above can produce more than one result. For example, a query such as charger un navigateur would produce load a browser, but there is also another option in French, which is appeler un navigateur. The two options should be retrieved based on the fact that both collocations are encoded with the LF Prepar1Fact0 (“prepare a key word to operate”). The same applies to other collocations with browser and password, as shown in Figure 8. The establishment of equivalence links between collocations can be extended to the display of entries. When looking at the entry for attachment listing all the

© 2009. John Benjamins Publishing Company All rights reserved

Combining the semantics of collocations with situation-driven search paths 275



Query: load a browser or run a browser load a browser launch a browser run a browser

appeler un navigateur charger un navigateur executer un navigateur Lancer un navigateur

Prepar1Fact0 Caus1Fact0

Query: Create a password create a password define a password select a password set a password

créer un mot de passe définir un mot de passe

Caus1Func0

Figure 8.  Translations of collocations with browser and password

collocations in which the English term can be found, a user can decide to visualize the translation of a specific collocation. The system will then (following steps similar to those given above) provide the translation (as illustrated in Figure 9). attachment1, n

Status: 2

Lexical relations That contains a virus The sender puts an a. in the email The sender removes an a. from the em ail The sender sends the file to the recipient as an a. The recipient starts using an a.

infected 1 ~ add an ~ to ... delete 1 an ~ send ... as an ~ open 1 an ~

FRENCH FRENCH FRENCH FRENCH FRENCH

Step 1: Locate the entry with the equivalent term: pièce jointe Step 2: Locate the specific lexical unit entered by the user: send Step 3: Extract the lexical function(s) that encode the relationship between send and attachment: Labreal123 Step 4: Locate the same lexical function(s) in the record for pièce jointe. send ... as an ~

Labreal123

envoyer … en attachement transmettre … en attachement

Figure 9.  Extracting equivalent collocations from entry display

Of course the process can be extended to other languages provided that equivalent terms are described in entries for other languages and that collocations are encoded with lexical functions. 5.2 Search path 2: production in L2 The second access method is designed to help users access specific collocations in records where very long lists are provided (for example, in the entry for the French

© 2009. John Benjamins Publishing Company All rights reserved

276 Marie-Claude L’Homme and Patrick Leroyer

term fichier in which, as was said above, nearly 150 relationships are listed). The general idea is to break down these lists in order to avoid reading long portions before being able to locate a collocate that expresses a given meaning. The strategy described in this section is based on that proposed in Jousse et al. (2008) who suggested a “thematic” approach to the browsing of the contents of a general language lexical database, the Dicouèbe (Jousse and Polguère 2005). The Dicouèbe lists paradigmatically and syntagmatically related lexical units and encodes them by means of lexical functions (also, the Dicouèbe provides a system of natural language paraphrases — based on Polguère 2003 — for explaining lexical functions). According to the thematic approach, the user can be guided through the network of relationships a lexical unit has with other lexical units using a limited number of paraphrases and can then be led to more specific questions in a step by step manner. The approach is suggested for all lexical relationships (and is not limited to collocates). Our method bears some similarities with that of Jousse et al. (2008). It also includes a series of generic paraphrases (based on Polguère 2003), but these are adapted to the fields of computing and the Internet and would be listed in a menu. A first series of paraphrases represents generic senses that are recurrent in the fields of computing and the Internet (such as “use/stop using N”, “create/delete N”, “have N”, “put N somewhere/remove N from somewhere”, “N functions/stops functioning”, etc.). For example, a user looking for specific collocates of file can be prompted with the options illustrated in Figure 10. Users can chose to visualize all lexical relations or, rather, access a menu in which paraphrases (leading to the display of specific sets of collocations) would appear. file, n. … Lexical relations

Display all

Find a word that expresses a specific meaning

Combined with units that mean: “create/destroy a file” “place a file somewhere/remove a file from somewhere” “use/stop using a file” “sort of file” Others

Figure 10.  Search for specific collocates of file

© 2009. John Benjamins Publishing Company All rights reserved

Combining the semantics of collocations with situation-driven search paths 277



If “use/stop using a file” were selected, the following collocations can be extracted: Load a file Open a file Edit a file Close a file Share a file

The “use/stop using a file” paraphrase can simply generalize series of specific lexical functions, e.g., Reali (“use”), Labrealij (“use for something”), IncepReali(“start using”), etc. On the other hand, if the user selects the “place a file somewhere/remove a file from somewhere”, he/she will obtain the following results: copy a file install a file move a file save a file store a file download a file

We can even take this method one step further and provide users who want to a second series of paraphrases designed to break down the generic level into more specific senses (e.g., “use” can be broken down into “start using”, “prepare for using”, “use”, “use for something”, etc). Below are examples of the results users can obtain if they select “use” to view collocates of file and computer. Refine your query: use a file “Prepare for using” Ø “Start using” open a file, load a file “Use” edit a file “Use for something” Ø “Stop using” close a file Others share a file

Refine your query: use a computer “Prepare for using” configure a computer “Start using” boot, start a computer “Use” run a computer “Use for something” run something on a computer “Stop using” shut down a computer Others

© 2009. John Benjamins Publishing Company All rights reserved

278 Marie-Claude L’Homme and Patrick Leroyer

Since we are dealing with collocates within a specific subject field, it appears possible to make a finite list of general meanings conveyed by them. In the field of computing and the Internet, since we are dealing mostly with nouns denoting entities (program, computer, data, etc.), the most recurrent meanings are those listed above for verbal collocates (“use/stop using N”, “create/delete N”, “have N”, “put N somewhere/remove N from somewhere”, “N functions/stops functioning”). This was also observed by Cohen (1986) in her dictionary of collocations in the field of stock exchange: in this particular field, collocates express meanings such as “increase”, “decrease”, in a very regular manner. Of course, there will also be some collocates that will not enter in a very broad classification and thus, it will be important to allow users to be able to access them without having to specify a given meaning. Finally, this second search path could also be combined with the first one and provide links to equivalents of collocations in other languages. 6. Concluding remarks — new paths and improved dictionary management It has been shown that dictionary search paths — the dictionary user’s access to the data — should be functionally determined by the actual user situation. In the case of the DiCoInfo, these paths should include a direct access to the translation of collocations in the form of equivalents — the L1-L2 function — as well as an access to the meaning of collocation when producing a text. The paths we designed are based on an extended exploitation of lexical functions, which provide a fine-grained, systematic description of collocational resources of the specialized language of computing and the Internet. The L1-L2 translation path is based on a formal relationship helping the user to move from forms to interconnected meanings through encoded meaning descriptions. Search path 2 goes in the opposite direction. It is based on prototypical, frame descriptions of specific meanings associated to conventional forms in the relevant field of computing, and is designed in its access mode to help users move from meaning to form, which in linguistic description is not the standard direction. Nevertheless, an efficient specialized lexicographic tool should also allow access to data for users searching for the right wording corresponding to a fairly precise idea of the meaning they wish to express. In comparison with search path 1, search path 2 is therefore of a more tentative nature, and remains open for further developments and refinements. At present, it is mainly designed to help users access specific collocations in records where very long lists are provided (for example, in the entry for the French term fichier in which, as was said above, nearly 150 relationships are listed). The general idea is to

© 2009. John Benjamins Publishing Company All rights reserved



Combining the semantics of collocations with situation-driven search paths 279

break down these lists in order to avoid reading long portions before being able to locate a collocate that expresses a given meaning. In a dictionary management perspective, our methods are well adapted to the ongoing updating of online dictionary resources, which by definition are “work in progress”. In this kind of resource, information can be extracted by users the moment it is added to a record and properly encoded. In addition, using meaning categories — the mapping of meaning scenarios — proves to be particularly effective to establish equivalence relationships between collocations associated with different languages. The use of meaning categories can also be extended to other languages. If records in a third or a fourth language are added, there is no need to redesign access paths or restructure data records. Multilingual dictionary projects become true networks of monolingual dictionary resources in which equivalence relations between collocations do not need to be established directly. It would be interesting in future work to extend search paths to include paradigmatic relations, and thereby offer extended access to the data. We would also like to define computational strategies to break down more complex sequences of words that could be entered by users. Finally, it is envisaged to extend the methods to pedagogical applications. As already discussed in Jousse et al. (2008) and Leroyer et al. (Forthcoming) in connection with lexical databases and general language applications, the expressive power of a semantic encoding could easily be extended to language for special purposes (LSP) learning, allowing a more systematic use of the lexicographic resources in other user situations, i.e. those triggered by cognitive needs.

Acknowledgments This work was carried out during a research stay at the Centre for Lexicography of the Business School of Aarhus (Denmark) financed by an Otto Mønsted grant. We would like to thank the members of the Centre for providing useful insights about ways to improve the access to collocations in the terminological databases developed at the Observatoire de linguistique Sens-Texte (OLST), University of Montreal and to take into account specific needs users of these databases may have. The DiCoInfo is developed partly thanks to funding by the Fonds de recherche sur la société et la culture (Québec) and the Social Sciences and Humanities Research Council of Canada (SSHRC). Finally, we would like to thank two anonymous reviewers for their comments on a previous version of this article.

© 2009. John Benjamins Publishing Company All rights reserved

280 Marie-Claude L’Homme and Patrick Leroyer

Notes 1.  L1 being the mother language of the user and L2 being any foreign language (and not necessarily a specific one). 2.  We are not saying here that this is not an important issue. In fact, the word combinations selected inevitably have an impact on how they will be listed in reference works and then retrieved from them. We simply are not addressing this problem here. 3.  More details about how these dictionaries treat collocations are given in L’Homme (2009a). 4.  The electronic version of the DAFA is presented in Section 2.2. 5.  However, for those readers who would like to obtain more information on these resources, references and URLs are supplied. 6.  It is worth mentioning, however, that in another dictionary compiled by researchers of the Centre of Lexicography (responsible for the compilation of the DA), a dictionary of Law, a distinction was made between collocations, short phrases and longer phrases. 7.  This entry has not been completed yet (English entries are not as complete as French ones) as indicated by the number 2 (0 labels complete entries). 8.  Items that are underlined are displayed if the user clicks on them. 9.  In fact, distinctions between series of paradigmatic relations (related meanings, opposites, morphological relationships, etc.) and syntagmatic relationships are made when terminologists encode lexical relations, but cannot be visualized in the Web interface. This should be done in a future version of the DiCoInfo. 10.  In printed dictionaries (Leroyer et al. 1999; Meynard 2000), lexicographers have decided on alphabetical ordering. This solves only partly access problems. 11.  More sophisticated strategies could be defined to break down more complex sequences and establish correspondences with the contents of DiCoInfo entries, for example if a user entered edit the binary file, this sequence should be divided into edit a file and binary file. 12.  The methods have been tested on several examples but have not been programmed yet. Hence, some refinements will probably be necessary. 13.  Of course, in order to allow efficient retrieval, the system should provide for a search on synonyms (for instance, if the user entered email attachment instead of attachment, the system should look for this synonym in the synonym data field of the DiCoInfo). 14.  In the DiCoInfo, the data is encoded in an XML structure (developed in collaboration with Guy Lapalme from the Département d’informatique et de recherche opérationnelle of the Université de Montréal). Equivalents appear in the records of source terms along with a link to the record in which they are described. In the English entry for attachment

pièce jointe

1



15.  The Labreal123 lexical function represents the following basic sense: the first argument (defined as the sender of the attachment) uses the base to send the second argument (defined as a file) to the third argument (defined as the recipient). 16.  In the DiCoInfo, lexical relationships are encoded as follows: In the English entry for attachment







Labreal123



send 1 … as an ~



In the French entry for pièce jointe





Labreal123



envoyer 2 … en ~



References Bergenholtz, H. 2005. “Den usynlige elektroniske produktions- og korrekturordbog.” LexicoNordica 12–2005, 19–38. Bergenholtz, H. and S. Tarp (eds.). 1995. Manual of Specialised Lexicography. The preparation of specialised dictionaries. Amsterdam/Philadelphia: John Benjamins. Cohen, B. 1986. Lexique de cooccurrents. Bourse–conjoncture économique. Montréal: Linguatech. DA = Dictionary of Accounting. Nielsen, S. and L. Mourier. 2009. Centre for Lexicography, Aarhus School of Business, Aarhus University. http://www.regnskabsordbogen.dk/regn/gbdk/ gbdkregn.aspx. Accessed 7 July 2009. DAFA 2000 = Binon, J., S. Verlinde, J. Van Dyck and A. Bertels. 2000. Dictionnaire d’apprentissage du français des affaires. Dictionnaire de compréhension et de production de la langue des affaires. Paris: Didier. DAFA 2009 = online version of Binon, J., S. Verlinde, J. Van Dyck and A. Bertels. Dictionnaire d’apprentissage du français des affaires. Dictionnaire de compréhension et de production de la langue des affaires. www.kuleuven.be/ilt/blf. Accessed 16 June 2009.

© 2009. John Benjamins Publishing Company All rights reserved

282 Marie-Claude L’Homme and Patrick Leroyer DiCoInfo 2009 = L’Homme, M.-C. et al. Le Dictionnaire fondamental de l’informatique et de l’Internet. Université de Montréal: Observatoire de linguistique Sens-Texte. http://olst.ling. umontreal.ca/dicoinfo/. Accessed 7 July 2009. Hausmann, F.J. 1979. “Un dictionnaire des collocations est-il possible?” Travaux de linguistique et de littérature 17(1), 187–195. Heid, U. and G. Freibott. 1991. “Collocations dans une base de données terminologique et lexicale.” Meta 36(1), 77–91. Jousse, A.-L. and A. Polguère. 2005. Le DiCo et sa version DiCouèbe. Document descriptif et manuel d’utilisation. Université de Montréal: Département de linguistique et de traduction. Jousse, A.-L., A. Polguère and O. Tremblay. 2008. “Du dictionnaire au site lexical pour l’enseignement/apprentissage du vocabulaire.” In Grossmann, F. and S. Plane (eds.). Lexique et production verbale. Vers une meilleure intégration des apprentissages lexicaux, coll. “Éducation et didactiques”. 141–157.Villeneuve d’Ascq: Presses Universitaires du Septentrion. L’Homme, M.-C. 2009a. [Forthcoming]. “A Methodology for Describing Collocations in a Specialized Dictionary.” L’Homme, M.-C. 2009b. [Forthcoming]. “Designing Terminological Dictionaries for Learners based on Lexical Semantics: The representation of actants.” In Fuertes-Olivera, P. (ed.). Festschrift in honour of Professor Alcaraz-Varo. Lerat, P. 2008. “Restrictions paradigmatiques et traduction de schémas d’arguments.” Meta 53(2), 434–442. Leroyer, P. 2001. Les exemples textuels unificateurs : fondements et applications pour le dictionnaire de traduction technique. Ph.D. Thesis presented at the Aarhus School of Business: Denmark. Leroyer, P. 2006. “Dealing with phraseology in business dictionaries: focus on functions — not phrases.” Linguistik Online 27 (2), 183–194. Leroyer, P., L. Kruse, É. Danino, A.M. Kiener and S. Tarp. 1999. Dansk-fransk erhvervsordbog. Viborg: Systimes erhvervsordbøger. Leroyer, P., J. Binon and S. Verlinde [in print]. “La lexicographie d’apprentissage française au tournant du troisième millénaire : le couple FLM/FLE(S) entre tradition et innovation.” Jahrbuch Lexicographica 25. Meynard, I. 2000. Internet. Répertoire bilingue de combinaisons lexicales spécialisées françaisanglais. Montréal: Linguatech. Mel’čuk, I (1998). “Collocations and Lexical Functions.” In Cowie, A.P. (ed.). Phraseology. Theory, Analysis, and Applications. 23–53. Oxford: Clarendon Press, 23–53. Mel’čuk, I., A. Clas and A. Polguère. 1995. Introduction à la lexicologie explicative et combinatoire. Louvain-la-Neuve (Belgique): Duculot/Aupelf — UREF. Mel’čuk, I., A. Clas et al.1984–1999. Dictionnaire explicatif et combinatoire du français contemporain. Montréal: Presses de l’Université de Montréal. Nielsen, S. 2006. “Monolingual accounting dictionaries for EFL text production.” Ibèrica 12, 43–64. Nielsen, S. and L. Mourier. 2007. “Design of a function-based internet accounting dictionary.” In Gottlieb, H. and J. E. Mogensen (eds.). Dictionary Visions, Research and Practice. 119–135. Amsterdam/Philadelphia: John Benjamins. Pedersen, J. 1998. “Kollokationelle og syntaktiske angivelser i en teknisk oversættelses- og produktionsordbog for fremmedsprogsbrugere.” LexicoNordica 6, 53–65.

© 2009. John Benjamins Publishing Company All rights reserved



Combining the semantics of collocations with situation-driven search paths 283

Polguère A. 2003. “Collocations et fonctions lexicales : pour un modèle d’apprentissage.” In Grossmann, F. and A. Tutin (eds.). Les Collocations. Analyse et traitement. 117–133. Amsterdam: De Werelt. Sager, J.C. 1990. A Practical Course in Terminology Processing. Amsterdam/Philadelphia: John Benjamins. Schneider, F. 1998. Studien zur kontextuellen Fachlexikographie. Das deutsch-französische Wörterbuch der Rechnungslegung. Lexicographica/Series major 83. Tübingen: Niemeyer Verlag. Tarp, S. 2008. Lexicography in the borderland between knowledge and non-knowledge : General lexicographical theory with particular focus on learner’s lexicography. Tübingen: Max Niemeyer Verlag. Termium = Termium Plus®. http://www.termiumplus.gc.ca/site. Accessed 15 February 2008. Wanner, L (ed.). 1996. Lexical Functions in Lexicography and Natural Language Processing. Amsterdam/Philadelphia: Benjamins Academic Publishers.

Authors’ address Marie-Claude L’Homme Observatoire de linguistique Sens-Texte (OLST) Université de Montréal C.P. 6128, succ. Centre-ville Montréal (Québec) H3C 3J7 [email protected]

Patrick Leroyer Department of Language and Business Communication Centre for Lexicography Aarhus School of Business, Aarhus University Fuglesangs Allé 4 DK-8210 Aarhus V [email protected]

About the authors Marie-Claude L’Homme is full professor at the Department of Linguistics and Translation of the University of Montreal where she teaches terminology and computer tools for translators. She is also the director of the Observatoire de linguistique Sens-Texte (OLST), a research group concerned with various aspects of the lexicon (general and specialized). Her main research interests are lexical semantics applied to terminology and corpus-based terminology. She develops along with a team of researchers in linguistics, terminology and computer science, terminological databases in the fields of computing, the Internet and the environment. Patrick Leroyer, PhD, is Associate Professor of French and Danish LSP and business communication at the Aarhus School of Business, University of Aarhus and is attached to the Centre for Lexicography. His current research is on theoretical lexicography in a broad, functional perspective, and includes the development of specific theories for data-access, -selection and -presentation in connection with communicative and cognitive functions in lexicographic information tools. He has published within the fields of LSP and business communication, terminography and lexicography.

© 2009. John Benjamins Publishing Company All rights reserved

© 2009. John Benjamins Publishing Company All rights reserved