Best practices for structuring a thesaurus
Domcumentary languages are controlled or standardized vocabularies, whose purpose is to respond to the equivocation of natural language and the problems posed by polysemy and homonymy. They are used to index and describe the cntent of a resource and are intended to improve documentary research.
There are three main types of document languages: classification schemes, authority or subject heading list and thesauri. Classifications, such as the Dewey Decimal Classification, propose a hierarchical organization of concepts. Authority lists or headings are standardized terms that must be used in cataloguing or indexing to describe the content of documents. We can cite Rameau (Répertoire d’autorité-matière encyclopédique et alphabétique unifié), the subject indexing language used by the Bibliothèque nationale de France. Finally, a thesaurus structures a controlled vocabulary by describing the set of semantic relations that link concepts.
The use of controlled vocabularies is a recommendation of the Agence Nationale de la Recherche for the evaluation of data quality in data management plans. Similary, according to the FAIR principles, ‘to be FAIR, data must be described using a controlled vocabulary that is itself FAIR compliant. It must be easy to find with a persistent and unique identifier, documented and machine-readable.
The structure of a thesaurus
The construction of a thesaurus follows ISO standards (ISO 25964-1:2011 and ISO 25964-2:2013).
The ISO 25964 standard distinguishes between concepts, as a ‘mental representation’, and terms, as a linguistic expression of a concept in a given language.
A concept is expressed by a preferred or descriptor term and non-descriptor terms. It has hierarchical and associative relationships with other concepts. It is aligned with concepts in other thesauri through equivalence relationships. It can be defined and documented by differents types of notes.
- Equivalence relationships between the preferred term and its synonyms
– The preferred term
In a knowledge organization system, the preferred (or retained, preferred) term is used to refer unambiguously to a concept. It is the descriptor of an indexing and the label appears in the display.
⚠ A thesaurus allows only one descriptor term per concept and per language.
⚠ The same descriptor term cannot designate several concepts. In case of homonymy, the scope of the term should be disambiguated by adding a qualifier in parenthesis:
⚠ A single term cannot be both a descriptor and a non-descriptor.
The descriptor term and the non-descriptor (or rejected, or non-preferred) terms that represent the same concept are related to each other by an equivalence relationship. It allows the control of synonyms and access to information.
According to the standards for constructing a thesaurus, are considered synonyms:
♦ Full names and abbreviations or acronyms: magnetic resonance imaging | IRM
♦ Common names and scientific names: spiders | Araneae
♦ Scientific and brand names: paracetamol | Efferalgan
♦ Terms of different linguistic origins: polychrome | multicolor
♦ Spelling variations: teacher | schoolmaster
♦ Ancient and modern appelations: $ exemple en anglais
Equivalence relationships also include quasi-synonyms. These are terms that are too specific to all appears as descriptors according to the domain addressed by the thesaurus.
UF : lacto-ovo vegetarian, lacto-vegetarian, ovo-vegetarian, semi-vegetarian, vegetarian.
Equivalence relationship can also include antonyms when they represent two aspects of the same concept, e.g. Literacy and Illiteracy.
- Hierarchical relationships
A hierarchical (or generic-specific or parent-child or broader-narrower) relationship must link, from the most general to the most precise, concepts of the same category: entities, activities, agents, actions, properties.
For example, the terms Museums (entities), Curators (agents) and Museography (action) do not belong to the same categories and therfore cannot be in a hierarchical relationship.
The ‘some/all’ test allows to check the validity of the hierarchical relationship:
General hierachical relationships can be subdivided into three categories:
– The generic relationship
Modeled after the genus-species relationship, the concept represented by the specific term is a particular type or species of the concept represented by the generic term.
– The partitive relationship
The hierarchical all-part relationship concernes:
Body systems and organs
BTP Digestive system
Disciplines and their specialties
– The instance relationship
This relation characterizes the link between a category or class designated by a common name and an instance of this category, represented by a proper name.
Each concept can be linked to several other concepts by hierarchical relationships.
BT Musical instruments
TS Keyboard instruments
TS String instruments
TS Stringed Instruments
BT Visual object
NT Geometric pattern
NT Cruciform geometric pattern
BT Symbolic objects
- Top Term
A Top Term is the preferred term that represents the highest level concept in the hierarchy to which it belongs. It is therefore located at the ‘root’ of the vocabulary and constitues an entry point into one of the branches of a thesaurus.
- Associative relationships
Associative relationships allow concepts that are not hierarchically linked but semantically linked to be brought together. To associate two concepts, the norm is that one must be a necessary element for the definition ot the other. For example,the concept of ‘Water supply’ is necessary for the definition of the definition of the concept of ‘pit’.
The associative relationships are reprocical. They make it possible to describe:
– Cause and effect:
NT Infectious diseases
– A whole and its part:
– An action and an agent or instrument:
– An action and the production of that action:
– An action and the object that undergoes that action:
– An action and the location of that action:
NT Primary schools
– A science and its object:
– An object or action and its property:
– An object or action and its role or purpose:
NT Water storage
– A raw material and its product:
– A process and its neutralizing agent:
– An object and its origin:
– An object or concept and a unit or mechanism of measurement:
– A profession and the person who practives that profession:
Associative relationships also allow for the accosiation of ‘brother’ concepts that share the same generic term and overlapping meanings (ships and boats), contrary concepts (marriage and divorce), and complementary concepts (teaching and learning).
⚠ Two concepts cannot be linked by both hierarchical and associative relationships.
|Type of relationships||Direction of the relationship||French terminology||Abreviations||English terminology||Abreviations|
|Equivalence relationship||⇒ from descriptor to non-descriptor||Employé pour||EP||Used for||UF|
|Êquivalence relationship||⇒ from non-descriptor to descriptor||Employer||E ou EM||Use|
|Hierarchical relationship||⇒ from specific term to generic term||Terme générique||TG||Broader Term||BT|
|Hierarchical relationship||⇒ from generic term to specific term||Terme spécifique||TS||Narrower Term||NT|
|Hierarchical generic relationship||⇒ from specific term to generic term||Terme générique (générique)||TGG||Broader term (generic)||BTG|
|Hierarchical generic relationship||⇒ from generic term to specific term||Terme spécifique (générique)||TSG||Narrower term (generic)||NTG|
|Hierarchical partitive relationship||⇒ from specific term to geneic term||Terme générique (partitif)||TGP||Broader term (partitive)||BTP|
|Hierarchical partitive relationship||⇒ from generic term to specific term||Terme spécifique (partitif)||TSP||Narrower term (partitive)||NTP|
|Hierarchical instance relationship||⇒ from specific term to generic term||Terme générique (instance)||TGI||Broader term (instance)||BTI|
|Hierarchical instance relationship||⇒ from generic term to specific term||Terme spécifique (instance)||TSI||Narrower term (instance)||NTI|
|Hierarchical relationship||Highest concept in the hierarchy||Terme de tête||TT||Top Term||TT|
|Associative relationship||⇒ from decriptor to another descriptor||Terme associé ; Voir aussi||TA ; VA||Related Term||NT|
The SKOS datamodel and its implementation in Opentheso
Thesauri can be described in SKOS (Simple Knowledge Organization System) format, whiwh is a W3C recommendation for representing document languages. This data model allows the publication, exchange and linking of structured vocabularies with Semantic Web technologies.
The classes of SKOS model and their equivalents in Opentheso:
|SKOS classes||Definition||Equivalent in Opentheso|
|skos:prefLabel||Descriptor term.||Preferred term|
|skos:altLabel||Non descriptor term.||Synonym|
|skos:hiddenLabel||The ‘hidden label’ allows to include wrong writtings of non-descriptor terms, integrated in the search but not visible to the user.||Hidden synonym|
|skos:broader||Broader concept.||Broader concept|
|skos:narrower||Narrower concept.||Narrower concept|
|skos:related||Related concept.||Related concept|
|skos:scopeNote||Scope, usage, use or explicative note: brief explanation on how to use a descriptor term.||Scope note|
|skos:historyNote||History note: describe a changment in the form or the signification of a concept.||History note|
|skos:editorialNote||Notes for the editorial managment of the thesaurus (comments).||Editorial note|
|skos:changeNote||Notes for the editorial managment of the thesaurus (documenting a change).||Change note|
|Alignment or equivalence with external standards|
|skos:exactMatch||Exact equivalence.||Exact match|
|skos:closeMatch||Close equivalence.||Close match|
|skos:broadMatch||Generic alignement links.||Broad match|
|skos:narrowMatch||Specific alignment links.||Narrow match|
|skos:relatedMatch||Associative alignment links.||Related match|
|Concept et collection of concepts|
|skos: collection||Collection of concepts.||Group|
|skos:notation||Code that identifies a concept.||Notation|
Plublishing your thesaurus
It is recommended that you write a presentation of your thesaurus on the home page. It is useful to specify the ‘metadata’ for the dissemination of a thesaurus – subject, context and date of creation, research project and term, institutions – and to indicate under which license it is published.
Aitchison, Jean, et Gilchrist, Alain, Construire un thésaurus. Manuel pratique, Paris, ADBS Éditions, 1992, 217 p.
Chichereau, Dominique, Contat ,Odile, Dégez, Danièle et al., « Les normes de conception, gestion et maintenance de thésaurus. Évolutions récentes et perspectives », Documentaliste-Sciences de l’Information, 2007/1 (Vol. 44), p. 66-74 [read online].
Dalbin, Sylvie, Yakovleff, Nathalie, Zysman, Hélène, ISO 25964-1- Thésaurus pour la recherche documentaire. Livre blanc, Saint-Denis, Afnor, 2013, 37 p.
Enssib, “Thésaurus”, 2013 [read online]
Glossaire des techniques de l’information, portail Joconde [read online].
Gouvernement du Québec, Politique d’édition du thésaurus de l’activité gouvernementale [read online].
Hudon, Michèle, « ISO 25964 : pour le développement, la gestion et l’interopérabilité des langages documentaires » Documentation et bibliothèques, vol. 58-3, 2012, p. 130–140 [read online].
Isaac, Antoine et Summers, Ed (ed.), SKOS : Simple Knowledge Organization System Primer. W3C Working Group Note 18 August 2009 [read online].
Keller, Loraine, Encadrer la réingénierie d’un thesaurus, mémoire pour obtenir le titre professionnel « Chef de projet en ingénierie documentaire », INTD, Conservatoire national des arts et métiers, Paris, 2013 [read online].
Serres, Alexandre, Introduction à l’indexation, URFIST Bretagne -Pays de Loire, 2003 [read online].
Alistair, Miles, Bechhofer, Sean, SKOS : Système simple d’organisation de connaissances. Référence. Recommandation W3C du 18 août 2009, trad. T. Francart [read online].
Zeng, Marcia Lei, Construction of Controlled Vocabularies, A Primer, 2005 [read online]