What it is
A term is not a word. ISO 1087 defines it as a designation representing a general concept by linguistic means, and the load in that definition is carried by the domain. A designation is a representation of a concept in a domain or subject, so termhood is a relation between a form and a field, not a property of the form itself. The same orthographic string can be a term in one subject field, a term for an unrelated concept in a second, and no term at all in general language.
Sager puts the structural consequence well: the lexicon of a special subject language reflects the organization of the discipline, tending to provide as many lexical units as there are conventionally established concepts, and restricting the reference of each unit to a well-defined region. General vocabulary does not work that way, which is why a specialized lexicon is thinner and sharper than a general one at the same size.
The second thing a translator needs from this record is that a term is not necessarily one word. Suonuuti's division is the plainest available and is freely consultable: a term may be a single word, a compound word, or a multiword expression. Nothing in the concept decides which of the three a given language will produce.
Where the authorities stand
The three-way structural division is standardized, and its numbering has moved. ISO 704's 2009 edition treats terms in clause 7.2; the 2022 edition has a subclause titled Types of terms at 7.3.2, whose text is not publicly readable. The vocabulary standard distinguishes a simple term, consisting of a single word or lexical unit, a compound term, which is a simple term that can still be split morphologically, and a complex term, consisting of more than one lexical unit. The distinction between the second and the third is the one translators get wrong, because it is morphological rather than orthographic: a compound is one lexical unit whatever the spelling does.
What is not settled, and what matters more in practice, is what holds a multiword term together. Cabezas-García's finding is that the semantic relation between the constituents of a complex nominal is normally left implicit, so the reader has to infer it. Air pollution and oil pollution are built identically and encode different relations, one naming what is polluted and the other what pollutes. Fine sand and carbonate sand look alike and encode an attribute in one case and a composition in the other. Longer strings add structural ambiguity on top: a night fashion show can be bracketed as a fashion show held at night or as a show about night fashion.
That is where a translator silently guesses wrong. The source language expresses the relation nowhere, the target language may require it to be expressed, and nothing flags the decision as a decision.
Consequences for translation
- Termhood is field-relative, so a term base entry without a subject field is not usable evidence. The same string can be standardized in one field and wrong in the next.
- Whether the target designation is one lexical unit or four is a fact about the target term stock, established before any translator arrives. It is not a stylistic expansion and should not be flagged as one.
- The unit count changes word counts and segment alignment, and it does so systematically rather than randomly. Pricing and memory leverage both feel it.
- With a multiword term, identify the internal relation before rendering it. If the source leaves it implicit and the target cannot, the translator is making a substantive decision and should record it.
Examples
EN>FR
- Source: mouse
- Target: moucheter
- Comment: The clearest possible case of field-relative termhood. In computer peripherals the English string is a standardized term for a pointing device; in zoology it designates a rodent; and in the outfitting of ships it is a verb, rendered in French as moucheter. One form, three fields, three concepts and two word classes, with nothing in the string to tell them apart.
EN>ES
- Source: wind turbine
- Target: aerogenerador
- Comment: Two lexical units against one. The Spanish designation is a single word built from combining forms, and the compression is a property of the Spanish term stock rather than a decision available to the translator. A reviser who reads the target as an omission has misread the direction of the fact.
FR>EN
- Source: gaz à effet de serre
- Target: greenhouse gas
- Comment: The mirror case, and the reason the previous one is not evidence of a general tendency. Here the French designation runs to four units and the English to two, for one concept in the same field. The unit count is a fact about each language's term stock at that node, and it does not point consistently in either direction.
References
- Sager, J. C. (1990). A Practical Course in Terminology Processing. Amsterdam: John Benjamins, at p. 19.
- Suonuuti, H. (2001). Guide to Terminology, 2nd edition. Nordterm 8. Helsinki: Tekniikan Sanastokeskus, section 3, p. 31.
- ISO 1087:2019. Terminology work and terminology science: Vocabulary. Geneva: International Organization for Standardization, entries 3.1.4, 3.4.1 and 3.4.2.
- ISO 704:2009. Terminology work: Principles and methods, 3rd edition. Geneva: International Organization for Standardization, clause 7.2.
- ISO 704:2022. Terminology work: Principles and methods, 4th edition. Geneva: International Organization for Standardization, clause 7.3.2.
- Cabezas-García, M. and León-Araúz, P. (2018). "Towards the Inference of Semantic Relations in Complex Nominals: a Pilot Study." Proceedings of LREC 2018. Miyazaki: ELRA.
- Cabezas-García, M. (2024). The Pragmatics of Multiword Terms: The Impact of Context. London: Routledge.
- Pearson, J. (1998). Terms in Context. Studies in Corpus Linguistics 1. Amsterdam: John Benjamins, at p. 12.