Concept systems and multidimensionality

What happens when two languages cut the same field along different lines. Also called concept diagrams or criteria of subdivision.

Reference/All terminology

Standardized Suonuuti (2001) ISO 704 (2009 and 2022) ISO 24156-1 (2014) Bowker (2022)

What it is

A concept system is what a set of concepts becomes once the relations between them are drawn. It is where terminology stops being a list of words with meanings attached and becomes a model of a field, and it is what a definition is written from.

Systems are built out of the relation types: generic, partitive, associative, and in practice mixed, since real fields rarely oblige one relation throughout. The drawing conventions are settled and freely consultable. Suonuuti uses a tree diagram for generic systems and a rake diagram for partitive ones, and adds a convention that repays the trouble: the criterion by which a superordinate has been subdivided can be written along the branch lines, so the diagram records not only what the subdivisions are but on what basis they were made.

Multidimensionality is what that convention exists to handle. A superordinate can be subdivided by more than one criterion, which is to say viewed from more than one dimension, and the resulting system is multidimensional. Concepts at the same level produced by the same criterion are coordinate; concepts produced by different criteria are not, however similar their position on the page looks. Wine divides by colour into red and white and by sugar content into dry and sweet, and the four are not four coordinate subtypes of wine.

Where the authorities stand

The framework is standardized, the numbering is not. Both editions of ISO 704 treat concept systems in clause 5.6, and the subclauses differ entirely: the 2009 edition covers their nature, their types and their modelling, the 2022 edition general matters, development and visualization, and types. Same number, different contents, so a citation without an edition is worse than useless.

The sentence carrying the concept is verifiable in the 2009 edition, where multidimensionality is introduced under concept relations rather than concept systems: a superordinate can be subdivided by more than one criterion, that is, viewed from more than one dimension, and the result is called multidimensional. The 2022 treatment is not publicly readable and should not be paraphrased.

One vocabulary point is worth fixing, because the two sets get mixed. ISO 704 uses multidimensional and dimension and does not use monodimensional; Suonuuti uses monodimensional and polydimensional and does not use multidimensional.

The 2022 edition also introduces concept models drawn according to ISO 24156-1, which sets out how a subset of UML symbols may be used, independently of their ordinary UML meaning, to represent concepts. That changes what a concept diagram may look like, not what it means.

Bowker supplies the history, and it is one of tooling as much as of theory. Because multidimensionality is complex and produces information overload, early terminologists concentrated on the most common dimension. The knowledge-based approaches of the 1990s, with multiple inheritance and generated diagrams, made several dimensions at once practical rather than merely conceivable.

Consequences for translation

Most equivalence failures that survive careful lexical work are failures of subdivision: the two languages have cut the same superordinate on different criteria, so the apparently corresponding subordinate is not the same node.

  • A subordinate term is comparable with another only if both were produced by the same criterion. Two terms at the same depth in two diagrams may have nothing in common.
  • Where the target subdivides and the source does not, the translator must supply a characteristic the source never expressed, and must find it in the document rather than the term base. Where the source subdivides and the target does not, the choice is between generalizing and describing.
  • Writing the criterion on the branch line, and on the term record, is the cheapest discipline available: it makes an unstated dimension visible before it becomes a mistranslation.
  • Mixed systems are the normal case, not a defect. A model that forces one relation type all the way down usually misclassifies something.

Examples

PL>EN

  • Source: udziały / akcje
  • Target: shares
  • Comment: Biel sets out the case: Polish company law subdivides on company type, so a private limited liability company has udziały and a public company has akcje, while the English terminology is unified regardless of company type. English shares is one node, Polish has two, and the criterion producing them is one English does not apply here. Translating into Polish, the company type has to be established before the term can be chosen.

EN>PL

  • Source: articles of association
  • Target: umowa spółki / statut
  • Comment: The same criterion again, at a different node in the same field. Biel records this pair alongside the shares pair: the private limited liability company has an umowa spółki and the public company a statut, where English uses articles of association for both, and English member likewise covers both wspólnik and akcjonariusz. One dimension is running through several nodes at once, which is what makes the divergence systematic rather than lexical.

References

  • ISO 704:2009. Terminology work: Principles and methods, 3rd edition. Geneva: International Organization for Standardization, clauses 5.5.2.1 and 5.6.
  • ISO 704:2022. Terminology work: Principles and methods, 4th edition. Geneva: International Organization for Standardization, clause 5.6 and Foreword.
  • ISO 24156-1:2014. Graphic notations for concept modelling in terminology work and its relationship with UML: Part 1, Guidelines for using UML notation in terminology work. Geneva: International Organization for Standardization.
  • Suonuuti, H. (2001). Guide to Terminology, 2nd edition. Nordterm 8. Helsinki: Tekniikan Sanastokeskus, section 1.4, pp. 14 to 19.
  • Bowker, L. (2022). "Multidimensionality." In P. Faber and M.-C. L'Homme (eds), Theoretical Perspectives on Terminology: Explaining terms, concepts and specialized knowledge, Terminology and Lexicography Research and Practice 23, 127 to 148. Amsterdam: John Benjamins.
  • Bowker, L. (1996). "Learning from Cognitive Science: Developing a New Approach to Classification in Terminology." Proceedings of EURALEX 1996, 781 to 787. Dublin: Dublin City University.
  • Biel, Ł. (2022). "Variation of legal terms in monolingual and multilingual contexts." In H. J. Kockaert and F. Steurs (eds), Handbook of Terminology Online. Amsterdam: John Benjamins, section 3.1.1.

Terminology that carries its evidence

Every Translyta delivery comes with a terminology report, so the equivalents behind a translation can be checked against the sources they came from.

Request a quote