Multilingual thesauri and ontologies in cross-language retrieval
Dagobert Soergel
College of Library and Information ServicesUniversity of MarylandCollege Park, MD [email protected]
AbstractThis paper sets forth a framework for the use of thesauri andontoiogies as knowledge bases in cross-language retrieval. Itprovides a general introduction to thesaurus functions, structure,and construction withparticular attention to the problems ofmultilingual thesauri. It proposes the creation of environmentsfor distributed collaborative knowledge base development as away to make large-scale knowledge-based systems more afford-able.
Outline
1 Introduction2 Thesaurus functions
2.1 User-centered/request-oriented indexing3 Thesaurus structure
3.1 Brief review of thesaurus structure principles3.2 Special issues in multilingual thesauri
4 Implementing thesaurus functions in retrieval4.1 Controlled vocabulary4.2 Free-text searching4.3 Knowledge-based support of searching
5 Thesaurus construction6 Affordable implementation of knowledge-based ap-
proaches
1 Introduction
A thesaurus is a’structure that manages the complexi-ties of terminology in language and provides conceptualrelationships, ideally through an embedded classifica-tion/ontology. This paper will
¯ give a tutorial on thesaurus functions and structure;
¯ present a concept retrieval perspective - conceptsbridge languages in retrieval;
¯ argue the perspective that cross-language retrieval isalso cross-cultural retrieval.
The paper covers thesaurus functions beyond retrieval;in retrieval, it considers any kind of object, not just text.
We start out with some definitions (Figures la and lb)and then give examples illustrating thesaurus problems andthesaurus structure (Figures 2 and 3).
Cross-language retrieval is the retrieval of any type ofobject (texts, images, products, etc.) composed or in-!dexed in one language (the target language) with a queryformulated in another language (the source language).There may be any number of source languages and anynumber of target languages. Queries can be written orspoken or constructed by selections from a menu pre-sented in the source language.
Text retrieval (broadly defined) is the retrieval of text,written or spoken. (While text retrieval has come to meanretrieval of written text, and speech retrieval retrieval ofspoken text, the broad meaning of text used here followsusage in linguistics).
Free-text retrieval is text retrieval based on the texttselfwithout index terms or other cues.
Fig. la. Definitions: Retrieval
A dictionary is a listing of words and phrases givinginformation such as spelling, morphology and part ofspeech, senses, definitions, usage, origin, and equivalentsin other languages (bi- or multilingual dictionary).
A thesaurus is a structure that manages the complexitiesof terminology in language and provides conceptual rela-tionships, ideally through a classification/ontology.
A thesaurus may specify descriptors authorized for in-dexing and searching. These descriptors then form acontrolled vocabulary (authority list, index language).
A monolingual thesaurus has terms from one language,a multilingual thesaurus from two or more languages.
A classification is a structure that organizes concepts intoa hierarchy, possibly in a scheme of facets. The termontology is ot~en used for a shallow classification ofbasic categories or a classification used in linguistics,data element definition, or knowledge management.
Fig. lb. Definitions: Thesaurus, etc.
197
From: AAAI Technical Report SS-97-05. Compilation copyright © 1997, AAAI (www.aaai.org). All rights reserved.
English German
simian Affemonkey niederer Affeape Menschenaffe
timepiece Uhrclock Wanduhr, Standuhr,
Turmuhrwall clock Wanduhrstanding clock Standuhrtower clock Turmuhr
watch Taschenuhr, Armbanduhrpocket watch Taschenuhrwrist watch Armbanduhr
alarm clock Wecker
blanket, rug, carpet Teppichblanket Betteppichrug, carpet Bodenteppich
rug (or carpet) loser Bodenteppichlong, narrow rug L/tufer
wall-to-wall carpet Teppichful3bodenhanging rug Wandteppich
Italics denotes terms created to express a concept notlexicalized in English or German, respectively.
Note that most English-German dictionaries wouldhave you believe that the German equivalent for"monkey" is "Affe", but that equivalence holds only insome contexts.
Another difficulty arises when two terms mean almostthe same thing but differ slightly in meaning or connota-tion, such as alcoholism in English and alcoholisme inFrench, or vegetable in English (which includes potatoes)and Gemf~se in German, which does not. If the difference:is big enough, one needs to introduce two separate con-cepts under a broader term; otherwise a scope note needsto clearly instruct indexers in all languages how the termis to be used so that the indexing stays, as far as possible,free from cultural bias or reflects multiple biases by as-signing several descriptors.
Fig. 2. Multilingual thesaurus problems
The example in Figure 2 illustrates the conceptual andterminological problems in aligning the vocabularies oftwo languages. Concepts lexicalized in one language maynot be lexicalized in the other and vice versa, creating sig-nificant problems for translation. The complexities of termcorrespondence are best managed with a conceptualapproach, establishing a concept interlingua, so to speak.
Figures 3a and 3b give a first illustration of thesaurusstructure, to be discussed more fully in Section 3. Theemphasis is on illustrating the concept-based approach tovocabulary management and retrieval, so the examples aredrawn from a thesaurus that epitomizes that approach,even though it is monolingual. The reader may want tolook at the sample pages from several multilingual thesaurigiven in the appendix.
Figure 3a presents an excerpt from a thesaurus hierar-chy. Through identifying the relevant facets (facet head-ings EF2, EF4, and EF6) and arranging the concepts withineach facet in meaningful order displaying the conceptrelationships, the hierarchy elucidates the conceptual struc-ture of the domain: The route of administration of drugscan be described by giving the intended scope of drug ac-tion, the method of administration, and the body site wherethe drug is administered. The last two facets have beencombined because they are strongly intertwined. Thehierarchy gives a logical arrangement for concepts within afacet, allowing the reader to form a clear mental image ofmethods available for administering drugs. The hierarchyalso allows for hierarchic query expansion (whethersearching with a controlled vocabulary or free-text).
Figure 3b gives examples of full thesaurus entries. Theentry for EF gives many synonyms that can be used forsynonym expansion of query terms. The RT cross-refer-ences suggest further descriptors that might be useful forsearching. The entries for EF2 give scope notes that care-fully define each concept. Thus, the thesaurus can serve asa reference. Juxtaposing the scope notes for hierarchicalneighbor concepts allows the indexer or searcher to pickthe right concept at the right hierarchical level.
Having established a general understanding of thesau-rus structure, we can now deal with the functions, struc-ture, and construction of thesauri in more detail.
198
EF route of administration
EF2EF2.2EF2.2.2EF2.2.4EF2.4
-- by scope of drug actiontopical and local administration
topical administrationlocal drug administration
systemic administration
EF4EF4.2EF4.2.2EF4.2.4EF4.4EF4.4.2EF4.4.4EF4.4.4.2EF4.4.4.2.2EF4.4.4.2.4EF4.4.4.4EF4.4.4.6EF4.4.6EF4.4.6.2EF4.4.6.4EF4.4.8EF4.6EF4.6.2EF4.6.2.2EF4.6.4EF4.6.6EF4.6.8EF4.6.10EF4.6.12EF4.6.14EF4.6.16EF4.6.16.2EF4.8
EF4.10
EF4.10
-- by method or body siteenteral administration
oral enteral administrationrectal enteral administration
mucosal administrationtransdermal administrationinhalation, smoking, sniffing
smoking.... smoking w/out inhalation.... smoking with inhalation
nasal administrationpulmonary administration
oral mucosal administrationbuccal administrationsublingual administration
rectal mucosal administrationparenteral administration
intravenous injectionintravenous infusion
intra-arterial injectionintraperitoneal administrationintracutaneous injectionadmin, through skin implantsubcutaneous injectionintramuscular injectionCNS injection
intrathecal injectionskin administration
(The full entry shows NarrowerTerm cross-references to themore specific methods involvingthe skin: EF4.4.2, EF4.6.8,EF4.6.10, and EF4.6.12)
oral administration(NT to EF4.2.2, EF4.4.4.2, andEF4.4.6)
rectal administration(NT to EF4.2.4 and EF4.4.8)
EF6EF6.2EF6.4
drug administration by self vs. othersself administration of drugsdrug administration by others
Fig. 3a. Excerpt from a thesaurus hierarchy
EF
EF2
EF2.2
EF2.2.2
EF2.2.4
EF2.4
route of administration
ST medication routeST method of delivery of drugs or foodST mode of substance administrationST route of drug applicationST route of drug entryST route of exposureBT +EEl2 pharmacokineticsRT +AA2 AOD useRT +BS AOD substance by route of admin.RT EE12.2e drug absorptionRT +EE14.4.8 drug effect by locationRT +HR drug therapyRT MD2.2.2.2 drug paraphernalia
route of admin, by scope of drug actionSN Use one of these descriptors in combination
with a descriptor from +EF4 route ofadmin, by method or body site.
topical and local administrationSN The application of a substance to a
localized area, chiefly for local effects
at this site.NT HU4.2 local anesthesiaRT GHI 0.2 chemical injury
topical administrationSN The application of a substance on
the surface of the skin or on a mu-
cous membrane (incl. the gastroin-testinal membrane) so that the sub-stance will take effect on the sur-face or on a localized layer underthe surface.For example, for the administrationof a decongestant spray, useEF2.2.2 topical administrationcombined with EF4.4.4.4 nasaladministration.
ST topical applicationlocal drug administration
SN The introduction of a substance intoa localized area of the skin or othertissue, as through injection.
NT EF4.6.4 intra-arterial injectionNT EF4.6.8 intracutaneous injectionNT +EF4.6.16 CNS injection
systemic administrationSN The introduction of a substance into
systemic circulation so that it is carried
to the site of effect.NT +EF4.6.2e intravenous injectionNT EF4.6.10 admin, through skin implantNT HU4.4 general anesthesiaRT +GH10.4 chemical poisoning
Fig. 3b. Examples of full thesaurus entries
199
2 Thesaurus functions
A thesaurus with its embedded classification/ontologyserves many functions, all of which are significantlyaffected by multilinguality. Our emphasis will be on the-saurus functions in retrieval (Figure 4a), but from a broad-er perspective one must not lose sight of the many otherfunctions a thesaurus and its embedded classifica-tion/ontology can serve (Figure 4b).
In information retrieval a thesaurus or a classification/ontology without the surrounding terminological structurecan be used in two scenarios:
(1) knowledge-based support of free-text searching(applicable only to written or spoken text, although the textcould point to another object, e.g., retrieving imagesthrough a free-text search of image captions or through asearch of the text portion of a movie);
(2) controlled vocabulary indexing and searching(applicable to any kind of retrieval).
The first two functions apply to either scenario. A usercan always profit from looking at a conceptual frameworkof the domain to clarify the search topic, and the thesaurusthen further assists in finding good search terms for theconcepts identified, Synonym expansion includes map-ping to terms from a different language.
Using a thesaurus as an indexing tool applies only tocontrolled vocabulary indexing. Indexing, the assignmentof a set of descriptors to a document or other object, can bemanual or automated. Of particular importance, so oftenoverlooked, is an approach to indexing that places theusers, their problems and questions squarely at the centerof attention: user-centered or request-oriented (or problem-oriented) indexing. The few empirical studies evaluatinguser-centered (as opposed to the commonly useddocument-centered) indexing show a positive effect onretrieval performance (Pejtersen 1983). This approach central in information retrieval; it will be discussed in Sec-tion 2.1.
To look at thesaurus functions more generally, we firstobserve that a thesaurus is a knowledge base of conceptsand terminology; other such knowledge bases are diction-aries and ontologies developed for AI applications, linguis-tic systems, or data element definition. Since these differ-ent types of knowledge bases -- though developed for dif-ferent purposes -- overlap greatly, it would be best to inte-grate them through a common access system (Soergel1996). The functions to be served by such a virtual inte-grated knowledge base of concepts and terminology arelisted in Figure 4b.
Knowledge-based support of searching(explicit assistance to the user or behind the scenes)
Menu trees
Guided conceptual analysis of a search topic
Browsing a hierarchy to identify search concepts
Mapping from query terms to descriptors used in oneor more databases or synonym expansion of queryterms for free-text searching
Hierarchical expansion of query terms
Meaningful arrangement of search results
Tool for indexing
vocabulary controluser-centered indexing
Fig. 4a. Thesaurus functions in information retrieval
2.1 User-centered/request-oriented indexing
As summarized in Figures 5a and 5b, user-centeredindexing involves analyzing actual and anticipated userqueries and interests and constructing a framework, a hier-archically structured controlled vocabulary, that includesthe concepts of interest to the users and thus communicatesthese interests to the indexers or an expert system that caninfer user-relevant concepts from text. The indexers thenbecome the "eyes and ears" of the users and index materi-als from the users’ perspective. The indexer uses the struc-tured list of user-relevant concepts as a checklist, applyingher understanding of a document (or other object) to judgeits relevance to any of these concepts. This process ensur-es that users will fred the documents that they themselveswould judge relevant upon examination.
Request-oriented indexing contrasts with document-oriented indexing, where the indexer simply expresseswhat the document is about or where simply the terms inthe text are used. But a document can be relevant for aconcept without being about the concept: a document ti-tled The percentage of children of blue-collar workers go-ing to college is not necessarily about intergenerationalsocial mobility, but a researcher interested in that topicwould surely like to fmd it, so it is relevant. Another ex-ample: Since users are interested in the biochemical basisof behavior and also in longitudinal studies, thesedescriptors are in the thesaurus. The indexer examines thedocument CSF studies on alcoholism and related behav-iors and fmds that it is relevant to both descriptors. Longi-tudinal is not mentioned in the document, but carefulexamination of the methods section reveals the concept.
200
¯ Provide a semantic road map to individual fieldsand the relationships among fields; relate conceptsto terms, and provide definitions, thus providingorientation and serving as a reference tool.
¯ Improve communication and learning generally:¯ Assist writers: suggest from a semantic field the
term that best conveys the intended meaning andconnotation.
¯ Assist readers in ascertaining the proper meaningof a term and placing it in context.
¯ Support learning through conceptual frameworks.¯ Support language learning and the development of:
instructional materials.
¯ Provide the conceptual basis for the design of goodresearch and practice.¯ Assist researchers and practitioners in exploring
the conceptual context of a research project, poli-cy, plan, or implementation project and in struc-turing the problem.
¯ Assist in the consistent definition of variables andmeasures for more comparable and cumulative re-search and evaluation results. Especially impor-tant for cross-national comparisons.
¯ Provide classification for action:¯ a classification of diseases for diagnosis,¯ of medical procedures for insurance billing,¯ of commodities for customs.
¯ Knowledge base to support information retrieval(Fig. ha)
¯ Ontology for data element definition. Data elementdictionary. Consider data processing systems in amultinational corporation
¯ Conceptual basis for knowledge-based systems.
¯ Do all this across multiple languages
¯ Mono-, bi-, or multilingual dictionary for humanuse. Dictionary/knowledge base for automated lan-guage processing - machine translation and naturallanguage understanding (data extraction, automaticabstracting/indexing).
Fig. 4b. Broader functions of a knowledge base ofconcepts and terminology
This kind of indexing is expensive, unless it can (to degree) be automated through a knowledge-based systemfor automated indexing. Is it worthwhile? The worth de-rived from improved performance depends on the use ofthe retrieval results.
Construct a classification/ontology (embedded in a the-saurus) based on actual and anticipated user queries andinterests.
Thus provide a conceptual framework that organizes userinterests and communicates them to indexers.
Index materials from users’ perspective:Add need-based retrieval clues beyond those available inthe document. Increase probability that a retrieval cluecorresponding to a query topic is available.
Index language as checklist.Indexing = judging relevance against user concepts.Relevance rather than aboutness
Implementation: Knowledgeable indexers or an expertsystem using syntactic & semantic analysis & inference.
Fig. 5a. User-centered / request-oriented indexing
Document
The drug was injected into the aorta
User concept: Systemic administration
Document:
The percentage of children of blue-collar workers goingto college
User concept: intergenerational social mobility
Document:
CSF studies on alcoholism and related behaviors
User concept: longitudinal study(Longitudinal not mentioned in the document; determin-ed through careful examination of the methods section.)
Fig. 5b. Request-oriented indexing. Examples
This perspective on indexing has implications forcross-language retrieval: The conceptual framework mustbe communicated in every participating language to allowa meeting of minds to take place, regardless of the lan-guages of the user and the indexer. This is particularlysalient in the context of indexing images with descriptorsthat capture imponderables, such as the mood of an image:One needs to make sure that, as far as possible, the termused by the indexer in one language communicates thesame mood as the term given to the user in another lan-guage for searching.
201
3 Thesaurus structure
At~er a brief review of general principles (3.1), we dis-cuss issues specific to multilingual thesauri (3.2).
3.1 Brief review of thesaurus structure principles
Thesaurus structure consists of the terminological structurethat relates terms to concepts (by establishing synonymrelationships and disambiguating homonyms) and concep-tual structure. We discuss each in turn.
Terminological structure (Figure 6)
Term
Alcoholism
Inheritance
Ultrasonic cardiography
Black
Afro-American
Pregnant adolescent
Controlling synonyms
Preferred synonym
Alcohol dependence
Heredity
Echocardiography
African American
African American
Pregnant teen
Disambiguating homonyms
administration 1 (management)
administration 2 (drugs)
Lltufer 1 (Sportier)
L/tufer 2 (Teppich)
Laufer 3 (Schach)
Engl.: runner (athlete)
Engl.: long, narrow rug
Engl.: bishop (chess)
discharge 1 (From hospital or program)German: Entlassung
discharge 2 (From organization or employment)Preferred synonym: DismissalGerman: Entlassung
discharge 3 (Medical symptom)German: Absonderung, Ausfluss
discharge 4 (into a river) German: Ausfluss
discharge 5 (Electrical)German: Entladung (which also means unloading)
Fig. 6. Terminological structure examples
The terminological structure is equally important incontrolled vocabulary and in free-text searching. In free-text searching, synonym expansion of query terms is im-portant, and homonym indicators can trigger a question tothe user on the intended meaning of the query term.
Conceptual structure
A well-developed conceptual structure is a sine quanon for user-centered indexing and is very useful for free-text retrieval as well. The two principles of conceptualstructure are facet analysis and hierarchy.
Facets. Semantic factoring or feature analysis
Semantic factoring means analyzing a concept into itsdefining components (elemental concepts or features).This gives rise to a concept frame with facet slots. SeeFigure 7 for examples.
liver cirrhosisPathologic process: inflammationBody system: liverCause: not specifiedSubstance/organism: not specified
alcoholic liver cirrhosisPathologic process: inflammationBody system: liverCause: chemically inducedSubstance/organism: alcohol
hepatitis APathologic process: inflammationBody system: liverCause: infectionSubstance/organism: hepatitis A virus
Fig. 7. Facet analysis examples
A facet groups concepts that fall under the same aspector feature in the definition of more complex concepts; itgroups all concepts that can be answers to a given ques-tion. In frame terminology: The facets listed above areslots in a disease frame; a facet groups all concepts that canserve as fillers in one slot.
Using elemental concepts as building blocks for con-structing compound concepts drastically reduces the num-ber of concepts in the thesaurus and thus leads to concep-tual economy. It also facilitates the search for general con-cepts, such as searching for the concept dependence, whichoccurs in the context of medicine, psychology, and socialrelations.
202
Facets can be defined at high or low levels in the hier-archy, as illustrated in Figure 8.
Top-level facets
organism
body partchemical substances by functionchemical substances by structure.
Low-level facets
route of administrationroute of administration by scope of drug action
(local/topical or systemic)route of administration by body partroute of administration by method of application
(injection, rubbing on, etc.)
liverliver tissue (hepatocyte, Kupffer cell, etc.)liver part (hepatic lobule, portal lobule, etc.)
Fig. 8. Facets at different hierarchical levels
Hierarchy
A sample hierarchy was presented in Fig. 3a. For an-other example, consider a search for a broad concept andthe more specific concepts that should be included in thequery as illustrated in Figure 9.
groups at high risk of drug use
suicidal or physically or mentally disabledpersons from unstable or low-cohesion familieschildren of alcoholic or other drug-abusing parents
SN Adult or still under agechildren of single teenage motherspersons subjected to abuse or neglect (now or past)
persons subjected to abuse/neglect by parentslatchkey children
persons subjected to abuse/neglect by spousesingle teenage mothersschool dropouts or those at risk of dropping outunemployed or in danger of being unemployedeconomically disadvantagedhomeless
runaway youthgateway drug userspersons engaged in violent or delinquent acts
Fig. 9. Hierarchy for searching
Uses of facet analysis and hierarchy
Through facet analysis and hierarchy building, the lexi-cographer often discovers concepts that are needed insearching or that enhance the logic of the concept hierar-chy; he then needs to create terms for these concepts. Ex-amples are traffic station as the semantic component com-mon to train station, bus station, harbor, and airport ordistinct distilled spirits (as the semantic component com-mon to gin, whiskey, cherry brandy, tequila, etc., the coun-terpart of the already lexicalized neutral distilled spirits),or analytic psychotherapy as an umbrella term for a host ofmethods (such as insight therapy, Gestalt therapy, and re-ality therapy) that all seek to assist patients in a personalityreconstruction through insight into their inner selves.
Fig. 10 lists the most important uses of facet analysisand hierarchy. These uses will be more fully discussed inSection 4.
Help to organize the concept space and establish conceptrelationships.
Assist the user in analyzing and clarifying a search prob-lem both in terms of the facets involved and the hierar-chical structure within each facet.
Facilitate the search for general concepts, such as
inflammation or
dependence (which occurs in the context of medicine,psychology, and social relations)
Hierarchic query term expansion
These functions are useful in both controlled vocabularyand free-text searching.
Fig. 10. Uses of facet analysis and hierarchy
3.2 Special issues in multilingual thesauri
A multilingual thesaurus for indexing and searchingwith a controlled vocabulary can be seen as a set ofmonolingual thesauri that all map to a common system ofconcepts. With a controlled vocabulary, indexing isconcept-based; cross-language retrieval is simply a matterof providing designations for these concepts in multiplelanguages so that queries can be written in multiple lan-guages. However, as the example in Fig. 2 illustrates, con-ceptual systems represented in the vocabulary of differentlanguages do not completely coincide.
203
The crux of the matter, then, is which concepts to in-elude. Ideally, the thesaurus should include all conceptsneeded in searching by any user in any of the source lan-guages. Language differences often also imply cultural andconceptual differences, more so in some fields than in oth-ers. We need to create a classification that includes allconcepts suggested by any of the languages. At a mini-mum this includes all relevant concepts lexicalized in atleast one of the source languages. Also~ different lan-guages often suggest different ways of classifying a do-main; the system needs to be hospitable to all of these. Theproblem that has bedeviled many developers ofmultilin-gual thesauri is that a concept lexicalized in one languagemay not be lexicalized in another and that the terms that doexist often vary slightly in meaning, possibly giving rise todifferent relationships. Starting from the misguided notionthat a thesaurus should include only concepts for whichthere is a term in the language and that term meanings can-not be adjusted for purposes of the thesaurus, they haddifficulty making the system of concepts the same for alllanguages. But, as we have seen, even in a monolingualthesaurus the lexicographer often discovers concepts need-ed in searching or to enhance the logic of the concepthierarchy and then needs to create terms for these con-cepts. In multilingual thesauri this necessity arises moreoften, particularly when different languages differ in thehierarchical levels at which they lexicalize concepts. Theprinciple proposed here is to establish a common concep-tual system, which may require an arduous, and expensive,process of negotiation, and then arrange for the terms in alllanguages to fit, giving proper definitions, of course.
It is clear that the problems discussed here and illus-trated in Fig. 2 and in Section 3.1 have major implicationsfor cross-language free-text searching: Each query termshould be mapped from the source language to its multipleequivalents in the target language; each of these equiva-lents may have other meanings in the target language, pre-senting potential problems for retrieval. The query termmay not have a precise equivalent in the target language;one may need to map to broader or narrower terms, distort-ing the meaning of the original query.
4 Implementing thesaurus functions in re-trieval systems
4.1 Controlled vocabulary
With a controlled vocabulary there is a defined set ofconcepts used in indexing and searching. Cross-languageretrieval simply means that the user should be able to use aterm in his own language to find the corresponding con-cept identifier that is used to retrieve documents (or what-ever the retrieval objects are). In the simplest system, thiscan be achieved through manual look-up in a thesaurusthat includes for each concept corresponding terms fromseveral languages and has an index for each language. Inmore sophisticated systems the mapping from term todescriptor would be done internally. As an example, con-sider a library catalog using the Library of Congress Sub-ject Headings, for which French and Spanish translationsare available. In the VTLS automated library system, eachsubject heading is identified by a number that is used in thedocument record. The authority file includes for each sub-ject heading the preferred term and any synonyms; thisinformation can be included in multiple languages. Onecould use this arrangement for assisting the user in findingsubject headings or automatic mapping of user terms tosubject headings as follows: Do a free-text search on au-thority records to find any subject heading for which eitherthe preferred term or any synonym contains the user’squery word or phrase in any language. Once appropriatesubject headings are found, they can be used to retrievedocuments.
Whenever the mapping from user terms to descriptorsis done "behind the scenes", transparent to the user, thesystem should ask the user for clarification whenever thequery word or phrase has multiple meanings and cannot bedisambiguated automatically. Beyond that, showing theuser the descriptor(s) the system came up with in their hi-erarchical context might improve the accuracy of the queryformulation and thus retrieval. The success of this type ofinteraction depends on the quality of the hierarchy and theinterface.
If the user has voice input available, one might even in-elude the spoken form of terms in the thesaurus to enablevoice input of query terms which would then be mapped tothe appropriate descriptors.
A cross-language retrieval system with controlledvocabulary must also support indexing of documents orother objects, that is the assignment of controlled vocabu-lary descriptors, in the various languages. For manualindexing, this is accomplished by having thesaurus ver-sions in each of the languages so that each indexer has aversion in her own language. But that is not enough. The
204
thesaurus version in each language must make sure that theindexer in that language fully understands the meaning of adescriptor that originated from another language; other-wise, the indexing of such a descriptor will not be consis-tent across the database.
Automated indexing with a controlled vocabulary, par-ticularly if it is to take a request-oriented slant, can be ac-complished with a knowledge base that (1) allows recogni-tion of important words and phrases (for spoken text thisrequires the inclusion of spoken forms) and allows forhomonym disambiguation and (2) gives mapping rulesthat lead from the (possibly weighted) set of words andphrases identified for a document to a set of descriptorsthat should be assigned.
Such mapping rules can take many forms. In their sim-plest form, they specify a direct mapping from text wordsor phrases to the appropriate descriptors for each word orphrase (and possibly even word or phrase combinations).To increase accuracy, the mapping can be made dependenton context (Hlava 97). A more complex mapping relies association strengths between terms (words and phrases)and descriptors. Broadly speaking, the association strengthbetween term T and descriptor D could be seen as the pre-dictive probability that the document containing term Tshould be indexed with descriptor D. Such associationstrengths can be computed from a training set of indexeddocuments. This is the approach often taken in automatedtext categorization, where often, but not always, the goal isto index each document by only one descriptor (assign it toone of a set of non-overlapping categories). An advancedversion of this approach is the use of "topic signatures",profiles consisting of a set of terms with weights; a docu-ment is assigned the topic if its terms match the topic sig-nature (Lin 1997). In effect, a topic signature is a querywhich identifies documents relevant to the topic.
As the foregoing discussion illustrates, the knowledgebase needed to support automated indexing is more com-plex than a thesaurus for manual indexing. It must includemore terms and term variants so that the words and phrasesimportant for indexing can be recognized in the text, and itmust include information needed for the disambiguation ofhomonyms (which often requires determining the part ofspeech of a text word).
For indexing and searching, a controlled-vocabularycross-language retrieval system can be seen as a set ofmonolingual systems, each of which maps the terms fromits language to a common system of concepts used in in-dexing and searching. For manual indexing and query for-mulation, this is accomplished through a multilingual the-saurus, which may in fact consist of multiple monolingualthesauri linked through common descriptor identifiers(such as Dewey Decimal class numbers). Automated in-dexing in cross-language text retrieval with texts in multi-
pie languages means mapping from each language to thecommon conceptual structure represented in the controlledvocabulary. The knowledge base component dealing withidentification of words and phrases for automated indexingcan be developed independently for each language. Map-ping rules that are entirely term-based can also be devel-oped independently for each language. However, somemapping rules, for example rules based on context or topicprofiles, may include conceptual elements that could beshared across languages.
There are a number of controlled-vocabulary cross-lan-guage retrieval systems based on manual indexing in use inbilingual or multilingual areas such as Switzerland, Bel-gium, Canada, and areas of the US with large Spanish-speaking populations; in international organizations, suchas the European Community; and in international collab-orative systems, such as AGRIS. These systems are basedon the Universal Decimal Classification, which has beentranslated into many languages (library of the ETH, Zu-rich); on the Library of Congress Subject Headings, whichhave been translated into French; on EUROVOC, an ECthesaurus in 9 languages; and AGROVOC, a thesaurus inthree languages created by translation from its originalEnglish-only version. There are a large number of thesaurithat either have been developed as multilingual thesauri orhave been translated into several languages.
4.2 Free-text searching
Cross-language free-text searching, finding texts in onelanguage that are relevant for a query formulated in an-other language without relying on controlled vocabularyindexing, is a more complex proposition. It requires thateach term in the query be mapped to a set of search termsin the language of the texts, possibly attaching weights ex-pressing the degree to which occurrence of a search term ina text would contribute to the relevance of the text to thequery term. To assist with this task, a thesaurus must in-clude the mapping information. If the thesaurus includesfine-grained definitions that deal with subtle differences ofmeaning, distance between such definitions can be used toderive term weights.
A major difficulty of this mapping is that a homonymused in the query gives rise to multiple translations, eachcorresponding to one of its meanings. The target termsmay in turn be homonyms in their language and thus re-trieve many irrelevant documents unless text terms are dis-ambiguated. (This problem exists in synonym expansionin one language as well but is exacerbated in cross-lan-guage text retrieval.) When the mapping goes to a termthat has multiple meanings, the specific meaning should beidentified, possibly in interaction with the user. For bestretrieval results the terms in the texts should also be disam-
205
biguated so that only documents that include the term inthe right sense score
The issue of homonymy in retrieval is not as straight-forward as it may seem at first glance (Sanderson 1994).First of all, quite a bit of disambiguation may occur "natu-rally", in that a given term may assume only one of itsmeanings in the specific domain of the collection andtherefore in the queries. Second, in a multi-componentquery, a document that includes a homonymous term fromthe first query component in a meaning other than that in-tended in the query is unlikely to also include a term fromanother query component, so excluding irrelevant docu-ments may not require disambiguation in either the queryor the texts. On the other hand, with single-concept queryto a general collection (such as the World Wide Web),disambiguation can be expected to have a beneficial effecton retrieval performance. Failing that, a system might beable to suggest to the user an additional query componentthat would separate out the documents that include thequery term but in a different meaning. Note that informa-tion extraction is much more dependent on homonymdisambiguation.
In any event, for best support of free-text retrieval athesaurus should flag homonyms, give their senses, andinclude rules for disambiguation.
The greater difficulty of free-text cross-language re-trieval stems in no small measure from the fact that onemust work with actual usage, while in controlled-vocabu-lary retrieval one can, to some extent, dictate usage.
4.3 Thesauri for knowledge-based search support
Whether searching is by controlled vocabulary or byfree text, it is often helpful to the user to browse a well-structured and well-displayed hierarchy of concepts,preferably with the option of including definitions. Amore sophisticated system may guide a User through afacet analysis of her topic. These aids provided by the sys-tem enable the user to form a better idea of her need and tolocate the most suitable descriptors or free-text searchterms. The guidance through facets and their hierarchicaldisplay must be available in the language of the user.These suggestions are based on the assumption that brows-ing a hierarchy is natural to most users and that users willappreciate the structure provided. This assumption restson the belief that people try to make sense of the world andthat guided facet analysis and browsing well-structuredhierarchies help them do so. There is anecdotal evidenceto support this assumption, but it needs to be investigatedby building prototype systems and studying users’ success(see, for example, Pollitt 1996).
This is one example of using a thesaurus as a knowl-edge base to make searching more successful. The assis-
tance provided does not require that the user be an expertin classification and thesauri. This is even more true for"behind-the-scenes" assistance. There is no need to teachusers about following a cross-reference from a synonym toa descriptor if the system searches for the descriptor auto-matically. There is no need to tell the user to look undernarrower terms also if the system can do a hierarchicallyexpanded search. There is no need to tell the user aboutstrategies of broadening the search if the system, in re-sponse to a user input that not enough was found, can sug-gest further descriptors to be searched based on cross-ref-erences in the thesaurus. Sophisticated retrieval softwarecan make the use of thesauri in retrieval independent of theuser’s knowledge and thereby can get much more mileageout of the investment in thesauri.
5 Thesaurus construction
Building a thesaurus, especially a multilingual thesau-rus, takes a lot of effort. Some term relationships can bederived by statistical analysis of term occurrence in cor-pora, but this will not result in the kind of well-structuredconceptual system described above. Developing such astructure requires intellectual effort.
A common method for thesaurus construction in a sin-gle language is to work bottom-up: One collects a list ofterms (words and phrases), preferably from search re-quests, but also from documents, free-term indexing, andother thesauri. These terms are then sorted into increas-ingly fine-grained groups, until a group contains only syn-onyms or terms that, for purposes of the thesaurus, can beconsidered synonyms. In this process at least some hom-onyms will be detected; they must be disambiguated intoseveral senses, each expressed by its own (possibly newlycoined) term having one meaning and being grouped ac-cordingly. A group of synonyms can be considered to rep-resent a concept; usually a preferred term to designate theconcept is selected, but some other concept identifier canbe used. A first rough hierarchy of concepts emerges fromthis process.
This is followed by conceptual analysis, especiallyfacet analysis at various levels, resulting in a well-struc-tured faceted hierarchy. Next, one needs to write defini-tions (scope notes), in the process of which one may re-think the hierarchy, and introduce relationships betweenconcepts that complement the hierarchy.
The development of a multilingual thesaurus is,naturally, an even more complex undertaking; the basicapproaches are summarized in Figure 12. The ideal way todevelop a multilingual thesaurus is to start from a pool ofterms in all covered languages and carry out the processwithout regard to the language of the terms. This will
206
bring together terms from different languages that have thesame meaning into one group. This process gives all lan-guages an equal chance to contribute concepts and conceptrelationships. It also forces a careful analysis of the mean-ing of each term in each language to determine the degreeof equivalence, making it possible to develop the fine-grained structure of definitions that has the potential ofproviding powerful support to free-text cross-languageretrieval.
Of course, this process would require a lexicographerknowledgeable in the subject matter of the thesaurus andfluent in all covered languages, not a very practical re-quirement. A more practical variation that still maintainsthe spirit of this approach is to start with two languagesand develop the conceptual structure -- a bi-lingual lexi-cographer is needed in any event. Definitions should bewritten in both languages. One would then work on a poolof terms in a third language and fit it into the structure,creating new concepts as necessary. This is not at all thesame as translating the thesaurus into the third language.This requires a lexicographer fluent in one of the startinglanguages and the third language. Following the sameprinciple, one can now add other languages.
The result of such a process is a conceptual system thatbrings the conceptual structures embedded in the differentlanguages under one roof, so to speak.
The most common approach to the construction of amultilingual thesaurus is to translate an existing monolin-gual thesaurus into one or more languages. But thisapproach is problematic: The original language and its vo-cabulary determine the conceptual structure, and one mere-ly looks for equivalent terms in the second language with-out covering its terminological richness. In somemultilingual thesauri, only one term in the target languagesis provided, making the thesaurus unsuitable for queryterm expansion in free-text searching.
In between is an approach in which one starts with amonolingual thesaurus as the center and fits terms fromone or more other languages into the structure of this cen-tral thesaurus without changing the concepts or the hierar-chy. EuroWordNet (Gillaranz 1997) takes an improvedvariation of this approach, working with the English Word-Net as its central thesaurus. In EuroWordNet, separate andindependent word nets are constructed in each language inparallel efforts, each identifying synonym sets in that lan-guage (A synset can be considered a concept). Each indiv-idual language project then independently maps itssysnsets to the corresponding WordNet synsets; no chan-ges are made to WordNet. In addition to identity, thismapping allows for hyponym and hypernym relationships,thus indicating that the concept identified in the languagebeing worked on is not included in WordNet, but giving atleast the hierarchical location. EuroWordNet also uses a
Requirements
Must cover all concepts of interest to the users in the var-ious languages, at a minimum all domain conceptslexicalized in any of the participating languages.Must accommodate hierarchical structures suggested bydifferent languages.
Approaches (by increasing complexity and quality)
(1) Start from monolingual thesaurus and translate. Thisapproach does not capture concepts lexicalized only inanother language and is biased to the conceptual structureunderlying the starting language. May not produce allsynonyms in the second language.
(2) Start from a monolingual thesaurus as the center.Collect terms from a second (third .... ) language and es-tablish correspondences of these terms to the central the-saurus. Suffers from similar bias toward the starting lan-guage as (1), but may cover more synonyms in the otherlanguages.
(3) Work with a central thesaurus as in (2), but aftercollecting terms from a second language first group theminto synsets, that is, derive concepts each of which isrepresented by a set of terms, and then map each conceptto the corresponding concept in the central thesaurus orindicate that the concept is new and give the nearestbroader or narrower concept in the central thesaurus.Note that the central thesaurus remains unchanged.
(4) As (2), but add concepts not in the starting thesaurus.This mitigates bias, but the central thesaurus now be-comes a moving target.
(5) Start from a pool of terms from all participatinglanguages and organize them into a conceptual frame-work, establishing term correspondence in the process.This approach results in a true "conceptual interlingua"not biased to any one language, but offering a home tomultiple conceptual perspectives. This approach requiresmost effort.
Fig. 12. Building multilingual thesauri
very weak variation of approach 5: The participants devel-oped a "top ontology", which presumably reflects and in-tegrates perspectives from their individual cultures. In ad-dition to being mapped to WordNet, the individual lan-guage synsets are also mapped to this top ontology.
207
Affordable implementation of knowledge-basedapproaches
¯ The effort needed for constructing and maintaining anyknowledge base, especially a well-structured multilingualthesaurus using method 4, is often forbidding, whence theattempts at constructing thesauri by statistical analysis ofcorpora. Fortunately, there is another way to reduce theeffort, often drastically: Capitalize on the intellectual.effortalready available in a multitude of existing thesauri anddictionaries by automaticaliy merging term relationshipsfrom many sources, as is done in UMLS (Unified MedicalLanguage System) or analyzing dictionary definitions toextract term relationships (Ahlswede 1988). Learn fromthe structure of text by creating hypotheses on the part ofspeech and semantic features of words during parsing(Sonnenberger .1995), or deriving term relationships fromuser queries. A further expansion of this approach calls forcollaborative development of thesauri and more compre-hensive databases of concepts and terms made possible bycomputer technology as proposed in Soergel 1996. Theseapproaches are summarized in Fig. 13.
Knowledge-based approaches require major investmentfor constructing the knowledge base.
Solutions
Use what is available (e.g. WordNet).
Reformat and integrate available sources into structuredknowledge bases.
Use machine learning techniques based on text or queryanalysis for building or adding to a knowledge base, per-haps followed by human editing.
Provide integrated access to multiple sources and an envi-ronment for distributed collaborative knowledge basedevelopment.
Fig. 13. Implementation of knowledge-basedapproaches
Conclusion
It was the intent of this paper to present a high-level re-view of the contribution knowledge-based systems canmake to cross-language retrieval, as exemplified in parti-cular through the structure and function of thesauri andontologies. Many of the ideas presented have been appliedin operational or experimental systems, even though em-pirical results need to be interpreted with caution (Soergel1994); others await application and testing
References
Ahlswede, Thomas, Martha Evens (1988). Generating relational lexicon from a machine-readable dictio-nary. International Journal of Lexicography 1, no. 3(Fall 1988): 214-237.
Ahlswede, Thomas, Martha Evens (1988). Parsing vs.text processing in the analysis of dictionary defini-tions. Proceedings of the Association for Computa-tional Linguistics (ACL) 26th Annual Meeting (June 7-10, 1988): 217-224.
Ahlswede, Thomas, Martha Evens, Kay Rossi, JudithMarkowitz (1986). Building lexicai database byparsing Webster’s Seventh New Collegiate Dictio-nary. In JOHANNESEN, Gayle (ed)., Advances inLexicology: Proceedings of the University of Waterloo(UW) Centre for the New Oxford English Dictionary2nd Annual Conference. (November 9-11, 1986): 65-78.
Anders, Monika (1986). Probleme der erarbeitung desmehrsprachigen thesaurus staat und reeht desISGI. Tell 1 [Problems encountered in the build-ing-up of the muitilingual thesaurus on state andlaw of the IISSS (Part 1)]. Informatik 33, no. 2(1986): 54-7.
Brodie, N.E. (1989) Canadians use a bilingual unioncatalog as an online public catalog. Library Trends37 no. 4 (1989): 414-431.
Buchinski, E.J., W.L. Newman, M.J. Dunn (1976). Theautomated subsystem at the National Library ofCanada. Journal of Library Automation 9 no. 4(1976): 279-298.
Canisius, P.W., P.M. Lietz (1979). Massnahmen zurOberwindung der spraehbarrieren in mehr-sprachigen dokumentationssystemen. Datenbasen,Datenbanken, Netzwerk. Hg. R. Kuhlen. MUnchen,1979:127-153.
Commission of the European Communities (1977). ThirdEuropean Congress on Information Systems and Net-works: Overcoming the Language Barrier,Luxembourg, 3-6 May 1977. 1. Mtinchen: VerlagDokumentation, 1977.
Belal Musstafa, A.A., T.S. Tengku Mohd and M. Yusoff(1995). SISDOM: multUingual do cument re-
208
trieval system. Asian Libraries 4~ no. 3 (September1995): 37-46.
Bennett, P.A., R.L. Johnson, John McNaught, JeanetePugh, J.C. Sager and Harold Somers (1986). Multilin-gual Aspects of Information Technology. Brookfield,VT: Gower Publishing Company, 1986.
Blake, P. (1992). The Men,USE System for multilinguaiassisted access to online databases. Online Review16, no. 3 (June 1992): 139-46.
Boisse, J. A. (1996). Serving multieuitural and multi-lingual populations in the libraries of the Universityof California. Resource Sharing and Information Net-works 11, no. I/2 (1996): 71-9.
Brendler, Gerhard (1970). Der mehrsprachige thesau-rus: Ein instrument zur rationalizierung desinformationflusses [The muitilinguai thesaurus: Aninstrument for increasing the efficiency of informa-tion flow]. Informatik 17, no. 4 (1970): 19-24.
Byrd, Roy J. (1987). Dictionary systems for office prac-tice. Automating the Lexicon: Research and Practice ina Multilingual Environment. Proceedins of the Linguis-tics Summer Institute Lexicon Workshop (July 1987):207-19.
Cousins, S.A. and R.J. Hartley (1994). Towards multilin-gual online public access catalogs. Libri (Interna-tional Library Review) 44, no. 1 (March 1994): 47-62.
Crofts, B.A. (1976). Problems Associations with the De-velopment and Maintenance of a Multilingual Thesau-rus. Report by Verina Horsnell, Chairperson. London,UK: British Library Research and Development, 1976.
Dijk, Marcel van (1966). Un thesaurus multilingue service de la corporation internationale [A multilin-gual thesaurus in the service of international eorpo-ration]. Bull De L’A.LD. 5, no. 4 (October 1966):85-7.
Djevalikian, Sonia (1980). Multilingual biblioservice.Bulletin-ABQ/QLA 21, no. 1 (January-April 1980):18-9.
Evens, Martha (I 989). Computer-readable dictionaries.Annual Review of lnformation Science and Technology(ASIS) 24 (1989): 86-118.
Food and Agriculture Organization of the United Nations(1995). Multilingual Soil Database~Food and Agricul-ture Organization of the United Nations, International
Soil Reference and Information Centre, Institute ofNatural Resources and Agro-Biology. Rome: FAO,1995.
Gilarranz, Julio, Julio Gonzalo, Felisa Verdejo (1997). Anapproach to conceptual text retrieval using theEuroWordNet Multilingual Semantic Database.Cross-Language Text and Speech Retrieval. StanfordUniversity, California: American Association for Arti-ficial Intelligence. (1997): 51-57.
Hlava, M.K. (1993) Machine-aided indexing (MAI) multilingual environment. Proceedings of the Four-teenth Mational Online Meeting (May 1993): 197-201.
Hlava, Marjorie, Richard Hainebach, Gerold Belonogov,Boris Kuznetsov (1997). Cross language retrieval English/Russian/French. Cross-Language Text andSpeech Retrieval. Stanford University, California:American Association for Artificial Intelligence.(1997): 66-72.
Horsnell, Verina(1976). WorkshoponMultilingualSys-terns: Report No. 5265 HC. West Yorkshire: BritishLibrary Research & Development Reports, 1976.
Humphreys, Betsy L., Catherine R. Selden (1997). UnifiedMedical Language System (UMLS): January 1986thorugh December 1996:280 citations. Current Bib-liographies in Medicine. Washington, D.C.: GPO(1997). [May 17, 1997: World Wide Web (WWW)Uniform Resource Locater (URL) http://www.nhn.nih.gov/pubs/cbm/umlscbm.html.]
Hutcheson, H.M. (1995) Preparation of multilingual vo-cabularies. Standardizing and Harmonizing Terminol-ogy: Theory and Practice. Philadelphia, PA: AmericanSociety for Testing and Materials. (1995): 102-114.
I SO R1149-1969e. Layout of Multilingual Classified Vo-cabularies.
209
ISO Standard 5964. 1985. Documentation Guidelines forthe Establishment and Development of MultilingualThesauri.
Johannesen, Gayle (ed). Advances in Lexicology: Proceed-ings of the University of Waterloo (UW) Centre for theNew Oxford English Dictionary 2nd Annual Confer-
ence (1986). Waterloo, Canada: UW Centre for theOxford English Dictionary, 1986.
Lavieter, L., J.A. Deschamps, und B. Felluga (1991). multilingual environmental thesaurus. TechnologyWork in Subject Fields (Proceedings). (October 12-14,1991): 27-44.
Layout of Multilingual Classified Vocabularies. 1st ed.Geneva: ISO, 1969.
Lin, Chin-Yew 1997 ([email protected], personal communic.)
Loth, K. (1993) AuJbau eines hierarchisch strukturiertenThesaurus. Juni 1993.
Loth, K. (1990). Aufbau eines eidgenOssischenBibliotheksthesauruus. ARBIDO-Revue 5 no. 4(1990): 111-113.
Loth, K. (1990) Gebrauchsanweisung zur ETHICS-Sachkatalogisierung. ZUrich: ETH=Bibliothek.Dezember 1990.
Loth, K. Computer searching of UDC numbers. Ency-clopedia of Library and Information Science 5 I.
McCallum, Sally and Monica Ertel (eds) (1994). Auto-mated Systems for Access to Multiliingual and Multi-script Library Materials: Proceedings of the SecondIFLA Satellite Meeting, Madrid, August 18-19, 1993.Germany: International Federation of Library Associa-tions and Institutions, 1994.
Muraszkiewics, M., H. Rybinski and W. Struk (1996).Software problems of merging muitilingual thesauri. Compatibility and Integration of Order Systems
(TIP/ISKO Meeting, Warsaw, 13-15 September, 1995).Warsaw: 1996: 58-67.
Paschenki, N.A., S. Ya Kalachkina, N.M. Matsak andV.A. Pigur ( 1981). Osnovnye printsipy sozdaniyamnogoyazychnykh informatsionno-poiskovykh [Ba-sic principles in producing multilingual thesauri].Nauchno Tekhnicheskaya Informatsiya 2, no. 5 (1981):17-9.
Pollitt, Steven 1997. http://www.hud.ac.uk/schools/cedar/-hibrowse
Ratzinger, M. (1991). Muitilingual product description(MPD), a European project. Terminology Work inSubject Fields (Proceedings) 45. (October 12-14,1991): 334-345.
Ritzler, C. (1990) Comparative study of PC-supportedthesaurus software. International Classification 17no. 3/4 (1990): S138-147.
Roelants, J.F.D. (1987). Le eatologage bilingueautomatise en Beigique. International Cataloging 16(1987): 16-18.
Rolland-Thomas, P., G. Mercure (1989). Subject accessin a bilingual online catalogue. Cataloging and Clas-sification Quarterly I0 no 1/3 (1989): 141-163.
Roulin, Corentin (1996). Bringing multilingual thesauritogether: A feasiblity study. Compatiblity andlnte-gration of Order Systems, Warsaw: (1996): 123-35.
Sanderson, Mark (1994). Word sense disambiguationand information retrieval. Proceedings of the Seven-teenth Annual International ACM-SIGIR Conferenceon Research and Development in Information Re-trieval, 3-6 July 1994. Dublin, Ireland. London:Springer-Verlag (1994): 142-151.
Schubert, K. (1985). Parameters for the design of anintermediate language for multilingual thesauri.Knowledge Organization 22, no. 3/4 (1995): 136-40.
Schuck, H.J, (1997). Sprrachliche aspekte bei der Clber-setzung von thesauri. Die Uberwindung der Sprach-barrieren. Bd. 1 Hg. M0nchen: Komission der Euro-paischen Gemeinschat~en, (1997): 399-414.
Semturs, F. (1997). Information retrieval from docu-ments in multilingual textual data banks. ThirdEu-ropean Congress on Information Systems and Net-works: Overcoming the Language Barrier:Luxembourg I (May 1977): 463-76.
Soergel, Dagobert (1974). Indexing languages and the the-sauri: Construction and maintenance. Los Angeles,CA: Melville; 1974. 632 p., 72 fig., ca 850 ref. (WileyInformation Science Series)
Soergel, Dagobert (1985). Organizing Information: Princi-ples of Data Base and Retrieval Systems. Los Angeles,CA: Academic Press; 1985.
210
Soergel, Dagobert (1994). Indexing and retrieval per-formance: The logical evidence. Journal of the Amer-ican Society for Information Science 45, no. 8 (1994):589-599.
Soergel, Dagobert. 1996. SemWeb: Proposal for anopen, miltifunctional mnitilinguai system for inte-grated access to knowledge base about concepts andterminology. Proceedings of the Fourth InternationalISKO Conference, 15-18 July 1996, Washington, D.C.Frankfurt/Main: Indeks ,Verlag: 1996. (Advances inKnowledge Organization, v.5, 165-173).
Sonnenberger, G. (1995). Der wit-interpreter: Eintextanalysesystem, das sein eigene wissenbasis durchdie analyse von texten erweitert (The wit-interpreter. text analysis system that adds to its own knowledgebase during.text analysis) (1995). Konstanz: Hartung-Gorre Verlag, 1995. (PhD. Thesis).
Stancikova, Pavia. (1996) International integrated data-base systems linked to multilingual thesauri cover-ing the field of environment and agriculture.Compatiblity and Integration in Order Systems(TIP/ISKO Proceedings). Warsaw, (1996): 163-9.
Still, J. M. (1993). Multilinguai and international data-bases. CD-ROM Professional 6, no. 6 (November1993): 72-6.
U.S. Department of Health & Human Services, NationalInstitute of Health, National Library of Medicine(NLM). (1993) Unified Medical Language System(UMLS) Semantic Network. (May 17, 1997: WorldWide Web Uniform Resource Locater ishttp://www.gsf.de/MEDWIS/pg__term/umls_net.html)
Velho Lopes, Roseane R. (1989). Automated access tomultilingual information: A Brazilian case study(Science and CIENTEC). Information Development 5(July 1989).
VTLS, Inc. (I 997) Library Automation and InformationManagement./May 17, 1997: World Wide Web(WWW) Uniform Resource Locatoer (URL)-http://www.vtls.com]
Walker, Donald E., Antonio Zampolli and NicolettaCalzolari (eds) (1995). Automating the Lexicon: Re-search and Practice in a Multilingual Environment.Oxford, New York: Oxford University Press, 1995.
Warwick, Susan. (1987). Automated lexical resources inEurope: A survey. Automating the Lexicon: Researchand Practice in a Multilingual Environment. Proceed-
ings of the Linguistics Summer Institute Lexicon Work-shop (July 1987): 329-63.
White, John (1988). Determination of iexieai-semanticrelations for multi-lingual terminology structures.Relational Models of the Lexicon, 326. Cambridge,UK: Cambridge University Press, 1988.
Wolff-Terroine, M. (1975). Multilingual systems (1975).Second European Congress on Information Systemsand Networks: Luxembourg (May 1975): 149-58.
Wolff-Terroine, Madeleine (1983). SABIR, Un logiciel degestion de thesaurus muitilingues" [SABIR, a pro-gram for managing multilingual thesauri]. MIDISTBulletin D’Information 2 (April 1983):
211
27. Ceramics and Glass
GermanM-15132
(En) Degussa.KERAMISCHES WOERTERBUCH. (Ger-En). 131p. $ 29.95
M7396(En) Hoffmann.
FACHWOERTERBUCH FLIER DIEGLASINDUSTRIE. DICTTIONARY FORTHE GLASS INDUSTRY. (Ger-En-Ger).1963. 160p. Bound. $125.00
ItalianM-7824 ?
(En) Vecchi. CERAMIC GLOSSARY. (lt-En). 1983.323p. $ 39.95
SpanishM-8393
(En) International Commission Glass.
DICTIONARY .OF GLASS-MAKING(Spanish supplement keyed to M-14381).
$ 45.00
SwedishM-8395
(En) International Commission Glass.
DICTIONARY OF GLASS-MAKING. (En-Sw). $ 45.00
Multilingual
M-8396International Commission on Glass.
DICTIONARY OF GLASS-MAKING(Dutch supplement keyed to M-8392).
$ 45.00
M-8339Clausen, H.
DICITONARY OF LACQUER TECH-NOLOGY. (Ger-En-Fr). 1980. 562p. Bound.
$125.00
M-15147CEC (Editor).
CERAMIC TILE DICTIONARY. (Ger-En-Sp-Fr-!t-Sw). 1970. 1,519p. Bound. $195.00
M-8394International Commission on Glass.
DICTIONARY OF GLASS-MAKING (Ital-ian supplement keyed to M-14381). $ 45.00
M-7749International Commission on Glass.
DICTIONARY OF GLASS-MAKING (Jap-anese supplement keyed to M-14381).
$ 45.00
M-8484International Commission on Glass.
DICTIONARY OF GLASS-MAKING (Pol-ish supplement keyed to M-14381). $ 45.00
M-15338International Commission on Glass.
DICTIONARY OF GLASS-MAKING (Por-tuguese supplement keyed to M-14381).
$ 45.00
M-7804International Commission on Glass.
DICTIONARY OF GLASS-MAKING (Rus-sian supplement keyed to M-14381). $ 45.00
Mol ho, D.!ctionary Catalogue
........ ~.-,;~ ........... . -. 2=’..
~:,:~,~,::: :: ../_ :::i!.i’:~: Chemistry 105¯ 7.,:- .-,~,~. t%.. ¯ ..... ~....::. ....... .
.... -. -~’-,"/ ..... :~.,::;~.,---~:.~.-z~ :,-,-M-8391 ...... ..,’~LM-.7687 ...... o-’=~:];~:~ :;
International Commission on Glass.." :~ ~.~’:,:,’.~.tEt"7.(En) S¢,hnclhnun~’Hans.DICTIONARY OF GLASS-MAKING ~4:v:¥DICWIONARY OF CHEMICAL PROCES-(Czech supplement keyed to M-14381). ": ~SES AND PRODUCTS. (En-Dut-En).
$ 45.00 ~1986. 120p. PlPbk. ~ $ 49.95
- 2 " ’~)~:;.~"~ ...... ’- ,:-::’" :’:’~M 83i~ternationaI Commission on Glass. ":~:~’~.ng//s? :.. ,:..g~.i:;,~.)~i.
DICTIONARY OF GLASS-MAKING. (En- _~.M-1875 ’ V:~.-.:i~:3~’r7 ,~ ~’:Fr-Ger). 1983. 402p. Ppbk. $ PNS :-2’ -;. (Ger) Schoeiifeld,
. . ......,~.~,THE CHEMIST’S ENGLISH. (En-Ger).M- 14381 -:-" :+, 2rid rev. ed. 1986. 173p, $ 22.95
International Commission on Glass. ..................ELSEVIER’S DICTIONARY OF GLASS- - :~" ~"-’ ......... ;-~:"~ ’,~-~!":~:.M-15297MAKING. (En-Fr-Ger). 1983. 402p. Bound. ’~?’.’7:~~ (Get) Weise, G. ’ .......
$ 250.00 ~ ENGLISCH FUER CI-IEMIKER. (En-..... Ger). 3rd ed. 1986. 256p. $ 29.95
M-8406 : " - .....International Commission on Glass~ :’./"--.’. M-15688 ............... ""
TERMINOLOGY OF DEFECTS IN ~ ............,~-~: ¯ Buckingham, John et al.GLASS. (En-Fr-Ger). 1976. 132p. $ 45.00 DICTIONARY OF ORGANOMETALLIC
¯ ¯ . COMPOUNDS. 3 vols. plus supplement.M-8031 " : ~ -1984 and 1986. 3,400p. Bound. $1,995.00
RTT 44. ¯ ....... ....GLOSSARY OF BUILDING GLASS. ..’ M-8022 ’ :: " - ¯(Nor-Dan-Ger-En-Finn-Fr-Sw). Each term is :- ....... Hamp~l, C.A. ’;...I’Ydefined in Norwegian. 1982. 64p. Ppbk. ~ GLOSSARY OFCHEMICAL
$ 29.95 ii. 1982. 320p. ~ ,:,oL~ . TE$39"95RMS_
M-14081. ’":.-M.7764 ": ".-~="..- ". ......TNC 27. . " ~’~ r’- Hawley, O.O. " ........
GLOSSARY OF ENAMELLING. (Sw-En- THE CONDENSED CHEMICAL DICTIO-Ger). Each term is defined in Swedish. 1957. ¯ NARY. 10th rev. ed. 1981.1,152p. Bound.104p. Ppbk. $ 29.95
.: i~ . ..... " " $ 95.00¯ . .- :.: ...... .::..: ../...-i.
M-15133Yon, Marguerite. ~" Finnish
DICTIONNAIRE ILLUSTRE MULTI- F-22467 " " ’.LINGUE DE LA CERAMIQUE DU ’. .... e ....... (Eta) Pihkala, L,and Pildmla, PROCHE-ORIENT ANCIEN. 1982. 312p. " ENGLANTILAIS-SUOMALAINEN LAB-
$ 99.50
28. Chemistry
ChineseF-94600
(En)ANGLO-CHINESE CHEMICAL DICTIO-NARY. 1,222p. Bound. $ 95.00
M-9248(En)
ENGLISH-CHINESE DICTIONARY OFCHEMISTRY AND CHEMICAL ENGI-NEERING. 1978. 1,458p. Bound. $ 95.00
M-9585(En)
ENGLISH-CHINESE DICTIONARY OFINDUSTRIAL CHEMISTRY. 1977. 81p.Ppbk. $19.95
CzechoslovakianM-15677
(Russ) Bubnikova, RUSKO-SESKY CHEMICKY SLOVNIK.(Russ-Cze). 1978. $ PNS
DutchM-15672
(En) Schnellman. Hans.DICTIONARY OF CHEMICAL ENGI-NEERING AND LABORATORY EQUIP-MENT. (En-Dut-En). 1986. 120p. Ppbk.
$ 49.95
ORATORIOJA PROSESSIALAN SAN-ASTO. (En-Finn). 1977. 184p. $ 49.95
French ..... :’:;" - .l=SSs / : " "(Dut) Schnellinann, Hans.
DICTIONNAIRE D’EQUIPMENT D’IN-DUSTRIE CHIMIQUE El" DE LABORA-TOIRE. (Fr-Dut-Fr). 1986. 120p. Ppbk.
$ 49.95
F-23291(DuO Schnelhnann, Hans.
DICTIONNAIRE DE PROCEDES ETPRODUITS CHIMIQUES. (Fr-Dut-Fr)..1986. 120p. Ppbk. _ ...... $ 49.95
M-7782. (En) Cornubert,
DICTIONNAIRE CHLMIQUE. (En-Fr).6th ed. 1979. 215p. Ppbk. $ 59.95
M-7158(En) Fromherz, Hans and King,
TERMINOLOGIE CHLMIOUE FRANCO-ANGLAIS. (Fr-En). 1968. 561p. $110.00
M-8007(En) Patterson, A.M.
FRENCH-ENGLISH DICTIONARY FORCHEMISTS. 2rid ed. 1954. 476p. Bound.
$ 49.95
M-8314(En) 8chnellmann, Hans.
DICTIONARY OF CHEMICAL PROCES-SES AND PRODUCTS. (En-Fr-En). 1986.120p. Ppbk. $ 49.95
212
.. International indexing andclassification bibliography
DETAILED CONTENTS OF PARTS I AND. II
Part I: BIBLIOGRAPHY OF UNIVERSAL CLASSIFICATION SYSTEMS
041 Library Classification Systems (by countries)041.8 Library Subject headings lists (by countries)
042042.1042. l-cs042.1-de042. l-en042. l-es042. l-fr042. l-hu042. l-it042. l-ja042. l-pl042. l-pt042. l-ru042. l-sk042.2042.3042.4042.5042.6042.7042.8042.9
Universal ~ Decimal Classification (UDC)Full Editions (by languages)
Czechoslovakian Full EditionGerman Full EditionEnglish Full EditionSpanish Full EditionFrench Full EditionHungarian Full EditionItalian Full EditionJapanese Full EditionPolish Full EditionPortuguese Full EditionRussian Full EditionSerbo-Croatian Full Edition
Medium EditionsAbridged EditionsProposals and Supplements to the UDCExtensions and Corrections to the UDCSpecialized UDC EditionsSpecial Purpose-Oriented UDC EditionsUDC Indexes and ThesauriConcordances, Correlations
043043.1043.3043.4043.5043.6043.7
Dewey Decimal Classification (DDC)DDC Full EditionsAbridged DDC EditionsProposals and Supplements to the DDCNotes and DecisionsSpecialized DDC EditionsSpecial Purpose-Oriented DDC-Editions
044044. l044.1.8044.3044.5044.7044.8
Library of Congress Classification (LCC)LCC Full Edition
Indexes to the LCCOutline of the LCCAdditions and Changes of the LCCSpecial Purpose-Oriented LCC EditionsLC Subject Headings (LCSH)
045 Bliss Bibliographic Classification (BBC)045.1 BBC Full Editions045.3 BBC Abridged Edition
046 Ranganathan Colon Classification (RCC)046.1 RCC Full Edition046.3 Outline of RCC046.6 Special RCC Schedules046.9 RCC Concordances and Correlations
2234777899
i0I0Iiii12131314181818
181919191920
202121212121
2122
22222226
213
SOLANACEAE (cont.)NT2 sol=mum quitoenesNT2 sel=mum tubarosum
NT1 withsniaNT2 withanla somnffera
It tublfloraeFr soleneceeeEa sol=macaaa
SOL&NINEBT1 glycoalkaloids
BT2 alkaloidsBT2 glycosidesBT3 carbohydrates
BT1 phytosterolsBT2 sterolsBT3 steroids
BT1 saponinsBT2 glycosides ."BT3 carbohydrates
Fr sol=mineEl sol=mine
SOLANUMBT1 sol=me.caseNT1 sol=mum aethloplcumNT1 sol=mum berthaultllNT1 sol=mum chaco=roseNT1 sol=mum hyporhodlumNT1 sol=mum khesl=mumNT1 sol=mum lacinlatumNT1 sol=mum malong=meNT1 sol=mum mudca~tumNTt sol=mum nlgrumNT1 sol=mum quito=mseNT1 sol=mum tubarosumFr sol=mumEl sol=mum
SOLANUM AETHIOPICUMBT1 sol=mumBT2 sol=maceae
Fr sol=mum aethlopicumEl sol=mum aethlopicum
SOLANUM BERTHAULTIIBT1 sol=mumBT2 sol=maceae
Fr sol=mum berthaultllES sol=mum berthaultll
SOLANUM CHACOENSEBT1 sol=mumB’F2 solanaceae
Fr sol=mum chacoenseEl sol=mum chacoense
SOLANUM HYPORNODIUMuf cocoasBT1 sol=mumBT2 sol=maceae
It fruit crops11 tropical fruitsFr sol=mum hyporhodiumEs sol=mum hypomodium
SOLANUM KHASIANUMBT1 sol=mumBT2 sol=macoae
rt drug plantsFr sol=mum khaal=mumE= sol=mum khaal=mum
SOLANUM LACINIATUMBTt sol=mumBT2 sol=masses
It drug pl=mtsFr sol=mum laclnlatumE= sol=mum laciniatum
solarium lyooperslcumUSE lyGoperalcon esculentum
$OLANUM MELONGENABT1 sol=mumBT2 sel=maceee
rt auberginesrt fruit vegetables
SOLANUM MELONGENA (cont.)It vegetable cropsFr sol=mum mal0ngenaE= sol=mum malongena "
SOL&NUM MURICATUMu/ pep/noBTI solariumBT2 solanaceae
It fruit cropstropical fruits
Fr sol=mum mudcatumEl sol=mum mudcatum
SOLANUM NIGRUMBT1 solariumB1"2 solanaceae
drug plantsFr sol=mum nigrum
solanum nlgrum
SOL&NUM QUrFOENSEuf lulouf naranjillaBT1 sol=mumBT2 sol=maceae
fruit cropstropical fruits
Pr sol=mum quitoenseEl sot=mum quitoense
SOLANUM TUBEROSUMBT1 solariumBT2 sol=macoae
potatoesroot vegetablesstarch cropsvegetable crops
Fr sol=mum tuberosumEl sol=mum tuberosum
solar activityUSE solar energy
SOLAR COLLECTORSBT1 equipment
heatingSOrer energy
Fr capteur solalmEs colsotoros solams
solar dryingUSE natural drying
SOLAR ENERGYuf solar activityuf solar radiationST1 renewable energyST2 energy sourcesB’r2 renewable resourcesBT3 natural resources
daylightgreenhouse effectsolar collectors
Fr energie solalmES energia solar
solar radiationUSE solar energy
solarimeters
USE aotlnometers
SOLE CROPPING(one crop grown -tone in purestands)uf single cropBT1 cropping systems
monocultureFr culture pureEl cultivo puro
sole, dover-., soles
sole, gray---, witch (fish)
AGROVOCso~, Im
--, soi~
~, to, bay.-~ , wit~(flsh)
8OLEABT1 pleumnectoidelBT2 bony fishesB’I3 fishes
rt solesFr sole=El asian,
SOLENST1 blvalviaB’r2 rnollusca
razor shellsFr solen
soten
SOLENOPOTESBT1 anoplum
BT2 phthlrapteminsects
BT4 aahropodaFr sol=mopotesEl solenopotes
SOLENOPSlSBT1 formlc~deeBT2 hymenoptsr=81":1 insects
BT4 arthropodaNT1 soleeopals geminalaNT1 sol=mopsis invictaNT1 solenopsis nchte~Fr esl=mopslsEs colenopsls
SOLENOPSIS GEMINATABT1 sol=mopslsB’r2 formic~desB’r3 hym=moptera
BT4 inesctsBT5 althropoda
Fr sol=mopsis geminateEl solenopsis geminate
SOLENOPSIS INVICTAST1 solenopsisBT2 formicideeBT3 hymsooptsra
BT4 inesctsBT5 arthropoda
Fr sol=mopsis invictaEl sol=mopals invicta
SOLENOPSIS RICNTERIBT1 sol=mopsisBT2 formlcidaeBT3 hymenoptera
BT4 inesctaBT5 =lhropoda
Fr esl=mopsis richtedEl solenopals dchtsd
SOLENOSTEMONBT1 labiatse
plecV=mthus esculentusIt root vegetables
starch cropsIt vegetable cropsFr sol=mo~temon
solenostemon
SOLES(iescaap group b-31)uf dover so/euf lascar (fish)uf lemon soleBT1 flatfish
BT2 saltwater fishesBT3 fishes
queneslia. soles
Pr soleEs lenguado
214
FRAN~AIS ITALIANO ENGLISH NEDERLANDS PORTUGUES
pr/s de rEAMA SAMA-paesi membri k4MS countries GASM-landen palsea EAMAkarhus Asrhus Aarhul Aarhus karhusenfant abandonn6 bambino ebbandoneto abandoned child vondeling c~n~ ebendonadaterra ebendonn6e terreno abbandoneto abandoned lend verleten land terra ebandonsdaaccord ABM eccordo ABM ABM Agreement anti-rsket-overeenkomst Acordo ABM
~r6 ’nlV
suppression des droits de douane
avortementAbruzzesabsent6isme
mejorit6 absolue
abolizione dei dazi doganali abolition of customs dgtJe= efachaffing van de douanerechten supressio dos direitos eduaneiros
eborto abortion abortus ebortoAbruzzo N) ruzzt Abruzzi Abruzoaessenteismo absenteeism absente~sme ebsentismo
meggioranza essoluta absolute majority absolute meerderheid maioria absoluteabstentionnisme estensioniamo ebltentlonlsm stemonthouding ebatencionismoAbou Dhabi At)u Dhabi N~u Ohabl Abu Dhabi Abu Dhabiebus de pouvoir abuso di potere abule of power machtsmisbruik ebuso de poderlibert6 de I’enseignement libert~ di insegnamento academic freedom vrijheid van onderwijs liberdede de ensinosocks ~i la profession eccesso erie professione access to a profession toegang tot her beroepslevenecesso ~ profissioacc~s ~ rinformation accesso alrinformazione access to Community Information toegang tot de communautaire ecesso i Jnforma~1o comun~ria
communsutaireacc~s =i l’6ducationacc6s ~ rinformationacc~s =i la justiceadh6sion ~, un accordsdh6sion =~ Is Communsut6
comunltariaaccesso alristruzioneeccesso alrinformazioneaccesso alia giustiziaedeaione ed un accordo
informatieaccess to education toegang tot het ondenvijs ecesso i educa~oaccess to Information toegang tot de informetie scesso i informa~oaccess to the coum toegang tot de rechtspleging ecesso i juati~accession toetreding tot een skkoord edes6o aum acordo
sdesione alia Comunit=i europea accession to the Community tostreding tot de Europeae ade~o ib ComunidadeGemeenschap
accident domestique infortunio domestico accident In the home ongeval thuis ecidente dom6sticopr6vention des accidents prevenzione degli infortuni accident prevention ongevallenpreventie preven~o de acidentescompte conto account rekening cuntscomptable contabile accountant boekhouder contabilistecomptabilit~ contabilit~ accounting boekhouding contebilidedeimputation comptable imputazione contabile accounting enW boekhoudkundige afschrijving imputac~o contebilisticasyst~me de comptabilit6 siatema di contabilit~ accounting system boekhoudsyateem ~tema contabilisticoacculturation acculturazione acculturation ecculturatie ecullura~oacids scido acid zuur /,cido
-EK
pluie scide pioggia acids acid rein zure regen chuva glcidaacoustique acustica acoustics geluidsleer sc0sticapays ACP ACP-paesi membri ACP countries ACS-lsnden peases ACPcomite des smbassadeurs ACP-CE comitato degli ambasciatori ACP-EC Committee of Comit6 van Ambassadeurs Comit6 dos Embaixadores ACP-CE
ACP-CE Ambassadors ACS-EGconvention ACP-CE convenzione ACP-CE ACP-EC Convention ACS-EG-conventie conven~o CEConseil des ministres ACP-CEinstitution ACP-CEAssembl6e paritaire ACP-CE
comit6 paritaire ACP-CEecquis communsutaire
consiglio dei ministri ACP-CEistituzione ACP-CEassembles paritetica ACP-CE
comitato paritetico ACP-CE"acquis" comunitario
ACP-EC Council of MlnleterzACP-EC InstitutionACP-EC Joint Assembly
ACP-EC Joint Committeeecquis communsuteire
Read van Ministers ACS-EGACS-EG-inatellingPeritaire Vergadering ACS-EG
peritair comit6 ACS.EGcommunautaire verworvenheden
Conselho de Ministros ACP-CEinstitui~o ACP-CEEAssemblela Parit~ria ACP-CE
Comit6 Paritirio ACP-CEscervo comunit6rio
acquisition de Is propri~t6recours contentieux administratif
recours contentieuxcommunautaire
recours du personnel
recours en ennuletion
acquisto della propriet~ricorso contenzioso amministrativo
ricorso contenzioso comunitario
rtcorso del personale
ricorso per ennullamento
acquisition of propertyaction brought before an
administrative courtaction brought before the EC
Court of Justiceaction by staff
action for annulment
eigendomsverkrijgingklacht inzake bestuurlijke
geschillenklacht inzake communautaire
geschillenberoep door het personeel
beroep tot nietigverklaring
equisi~o de propriedederecurao contencioso edministrativo
recurso contencioso comunit~do
recurso do pessosl
recurso de anulagiorecours en annulation CE
recours en carencerecours en manquement
programme d’actionrecours en responssbilite
administrative
ricorso per annullamanto CE
ricorso per carenzaricorso per insdempienza
programma d’azionericorso per accertamento di
responsebilit~ amministrativa
action for annulment of en ECdecision
action for failure to actaction for failure to fuffll an
obligationaction programmeaction to establish liability on the
part of an administration
vordering tot nietigverklering
beroep wegens onvermogenberoep wegens niet-nakomen
sctieprogrammavordering tot edministratieve
eansprakelijkheid
recurso de enula~o CE
ac~o por omisslioac~o por incumprimento
programs de ac~orecurso de responsabilidade
administrative
215
ESPAI~IOL
paise,= de la EAMAAarhushijo abandonadotierra abandonadaAcuerdo ARM
DANSK
A~,SM-landeArhus Amtfodadt barnforladt jord’ABM-affale
ENGLISH
A/I~S countrleeAmusabandoned childabandoned landABM Agreement
DEUI"SCH
AkSM-StaatenAarhusausgesetztes Kind.SozialbracheABM-Vertrag
supreai6n de los derechos deaduana
abortoAbruzosabsentismo
mayoria absoluta
ophaevelse af toldafgiffer
abortAbruzzlarbejdsforsemmelsa
absolut flertal
abolition of customs dutiel
abortionAbruzzlabsenteeism
absolute majoflty
Abschaffung der Z611e
AbtreibungAbruzzenFernbleiben yon der Arbeit
absolute MehrheitabstencionismoAbu Ohabiabuso de poderlibertad de ensefianzaacceso a ta profesi6n
manglende valgdeltagelseAbu Dhabiembedsmisbtugundervisningsfrihedadgsng til erhverv
abstantionlamAbu Ohabiabuse of poweracademic freedomaccen to e profe=don
WahlenthaltungAbu DhabiAmtsmiBbrauchFreiheit der LehreZugang zum Beruf
a~ox~AI~r~o0 Nrdpmx=rd=XPqml c~ouai=¢;ckcu~¢pi.¢ CK~¢i.6¢U~C;~p6oeao’q o¢ ¢’n~y~ckp.a
acceso a la informaciOncomunitaria
acceso a la educacibnacceso a la informacibnacceso a la justiciaadhesibn a un so, erda
adgang til EU.oplysninger
adgang til uddannelseadgang til informationadgang til retsv~esenettiltraedelse af en afraid
access to Community Information
acceu to educationaccess to Informationaccess to the court=accession
Zugang zuGemeinschaffsinformationen
Zugang zur BildungInformationszugangZugang zur RechtspflegeAbkommensbeitritt
’npbo’8~ml crqv KOcvO’nX~lwXqpo~obpqml
rrpba8r,~ a’rqv ,~K’nCZ[5,’U~wpbaSaml O’nlV r~Mlpoqoopiarrpbo6aml o~1 5zxmoo’Ovq~Tpoo’x~PTIOTI o¢ o’ul~uv[a
sdhesi6n a la Comunidad Europea
accidents dom~sticoprevenci6n de accidentescuentacontable
tHtra~delse af EF
ulykke i hjemmetforebyggelse at ulykkerregnskabregnskabsf~rer
accession to the Community
accident in the homeaccident preventionaccountaccountant
Beitritt zur Gemeinschaft
Unfall im HaushaltUnfallverhStungKontoBuchhalter
’npoo73bp’rlm10’nlV Evpo,-’noz[~Kotv6’rq’m
¢ruXfll~a~a a’ro ~i~=rrp6Xqqaq ’r~av ~ruXrl~r~vkoyapiaop6~;~oy,o-r~¢;
conlabilidadimputacion contablesistema de contabilidadaculturacibn~cido
ragnskabsv~senkonteringregnskabssystemtilpasning til andes kulturmiljosyre
accountingaccounting enWaccounting systemaccultur~onacid
BuchfOhrungVerbuchungKontenrahmenAkkulturationS~ure
kO~¢~nKfi
Xoy~o’rtx6 oOo’r~lp,¢=’rroXm~=x~l c~¢pOl~Oiu~o~0
Iluvia &tideacusticapaises ACPComit~ de Embajadores ACP-CE
convenio ACP-CE
sur regnakustikAVS-landeAmbasssderudvalget AVS-EF
AVS/EF*konvention
acid rainacousticsACP countriesACP-EC Committee of
Ambassador=ACP-EC Convention
saurer RegenAkustikAKP-StaatenBotschatterausschu8 AKP-EG
Abkommen AKP-EG
6{~vn 8poxh(~KOU(Tr.~KTIX~p¢~ AKEE~vrpo’r’~ rlp~oBz~v AKE-EK
o0pSao~ AKE-EKConsejo de Ministros ACP-CEinstitucion ACP-CEAs~mblea paritaria ACP-CE
comite pariterio ACP-CEacervo comunitario
Ministerr~det AVS-EFAVS/EF.institutionDen Paritetiske Forsamling
AVS-EFDet Paritetiske Udvalg AVS.EFgaeldende f~ellesskabsret
ACP.EC Council of MlnlstemACP.EC InstitutionACP.EC Joint Assembly
ACP.EC Joint Committeeacquls communautalre
Ministerrst AKP-EGAKP-EG-InstltutionParitdtische Versammlung AKP-EG
Paritdtischer Ausschu6 AKP-EGBesitzstand der Gemeinschaft
Zup.8OOXlO Y’noupyd~v AKE.EK6pyavo AKE-EKZuvkXcUOTI ’1m1¢;
EKwpoo’uwqml(; AKE-EKIo’o~pflc; Em’rpo’rr~ AKE-EKKOWO’r~K6 KCK’rqlJ.~VO
adquisici6n de la propisdedrecurso
contdncioso-administrativorecurso contencioso comunitario
rscurso del personal
recurso de anulaci6n
erhvervelse af ejendomforvaltningssager
faellesskabsretlige sager
klageadgang for ansatte
annullationssegsm~l
acquisition of propertyaction brought before an
administrative courtaction brought before the E¢
Court of Justiceaction by staff
action for annulment
Eigentumserwerb ~’rflml xupz6’nlra¢;Klage vor dem Verwaltungsgericht ~poo’qou~ a’rr, 51o=xq’n~.~
Klage in EG-Streitsachen ’npo=qouy~l cr’rllv KO|vo’rIKh&xa¢ooOvrl
Verwaltungsbeschwerde der ~po~q~u~ unc~Ikou x=r~,Bediensteten ’rrK 6~o{x~orl(;
Anfechtungsklage "npooqouYfl ~xup~O’l;~
recurso de anulaci6n CE
racurso par omisi6nrecurso par incumplimiento
programs de actuaci6nrecurso par responsabilidad
administrative
annullationss~gsm~((EF)
passivitetss~gsm~ltraktatbrudssag
handlingsprogramerstatningssogsm~l i
forvaltningssager
action for annulment of an ECdecision
action for failure to actaction for failure to Mill an
obligationaction programmeaction to establish liability on the
part of an administration
EG-Nichtigkeitsklage
Unt~tigkeitsklageKlage wegen Vertragsvertetzung
AktionsprogrammAmtshaftungsklage
~poo~pu~ ~KupGocu~; EK
~pooqou~ cwi wcxpczXcEq~¢=rrpoo~pUYfl crri ’nc(pc~B&aSl
’np6yp(x~pa (~pa~(;
216