Use definitions to agree scope
The same word can mean different things in different proposals. “Editing” might describe bilingual revision, target-only review or a change to the source. A buyer should therefore name the material checked and the outcome required. The ISO 17100 guide explains the standard’s production stages, while the definitions below provide a working vocabulary for the project agreement.
Tool-specific labels also need interpretation. A match percentage describes similarity according to a configured analysis, not a guarantee of accuracy or a fixed discount. The translation quote guide connects those categories to commercial scope. The glossary keeps the general concept separate from any provider’s charging policy.
Terminology
Translation terms
defined
- Source language
- The language of the material being translated. The source package should identify the authoritative version, since several language versions of a document may exist without all having equal approval status.
- Target language
- The language into which material is translated. A specification may also need a regional variant, audience and intended use; naming a language alone does not resolve every terminology or style decision.
- Locale
- A combination of language and relevant regional or cultural conventions used for a product or audience. A locale can affect formats, sorting and terminology, while the currency of individual records remains separate data.
- Translation
- Rendering written content in another language while preserving its meaning and meeting the agreed purpose. It involves choices about terminology, structure and register rather than substituting words one at a time.
- Localisation
- Adapting content or a product for a defined market’s language and conventions. It can include formats, typography and interface behaviour as well as translation; the project agreement should identify which aspects are included.
- Internationalisation (i18n)
- Designing a product so it can support different languages and regional conventions. Typical concerns include Unicode, flexible layouts, externalised strings and locale-aware formatting. It creates the technical conditions for localisation rather than supplying translated content itself.
- Transcreation
- Creative adaptation that preserves a communication goal or brand intent while allowing substantial changes to wording. The brief should identify the audience, constraints and approval criteria rather than expecting sentence-by-sentence correspondence with the source.
- CAT tool
- Computer-assisted translation software used by a linguist to work with segments, reference resources and protected formatting. CAT does not mean machine translation, although a CAT environment may optionally connect to machine-generated suggestions.
- Translation memory (TM)
- A database of source segments paired with target translations for retrieval and reuse. Its usefulness depends on language direction, domain, approval status and maintenance; a large memory can still contain outdated or unsuitable wording.
- Termbase
- A structured terminology resource containing concepts, terms and supporting information such as definitions or approval status. Unlike a translation memory, it normally manages terms and their relationships rather than storing whole translated sentence pairs.
- Glossary
- A list of terms and explanations or approved equivalents for a subject or project. A bilingual glossary is more useful when it includes context, definitions, prohibited alternatives and a named owner for terminology decisions.
- Style guide
- Instructions governing writing choices such as register, capitalisation, punctuation and product naming. A project should identify its current version and explain how it ranks against other references when instructions appear to conflict.
- TMX
- Translation Memory eXchange, an XML-based format for exchanging translation-memory data. An export transfers bilingual content, but compatibility should be tested where tags, metadata or segmentation rules differ between the originating and receiving tools.
- TBX
- TermBase eXchange, a format family for exchanging structured terminology data. The required dialect and fields should be agreed, since a simple spreadsheet and a concept-based termbase may represent terminology in different ways.
- XLIFF
- XML Localization Interchange File Format, used to carry localisable content through a workflow. Version, supported features and inline codes matter: two systems accepting XLIFF may still need a compatibility check before a project starts.
- Segment
- A unit handled by a translation tool, often a sentence or heading. Segmentation depends on rules and file structure, so a segment is not necessarily a complete thought or a visually separate paragraph.
- Alignment
- Matching an existing source document with its translated counterpart to create bilingual segment pairs. Alignment results need checking because differing structure, omissions or revisions can pair text that does not actually correspond.
- Concordance search
- Searching a translation memory for a word or phrase within stored segments. It helps locate earlier usage, but a retrieved example still needs assessment for current context, terminology status and source-version relevance.
- Fuzzy match
- A memory suggestion whose source resembles the current segment without matching it exactly under the tool’s rules. Differences can be small but significant, including changed quantities, negation or a different component name.
- 100% match
- A translation-memory result whose source segment matches the current source according to the analysis rules. It does not independently prove that the stored translation is correct, current or suitable for the surrounding document.
- ICE match
- An in-context exact match: exact source text with matching context under a tool’s defined rules. Context criteria and commercial treatment vary, so the analysis settings should accompany any quote that relies on this category.
- Repetition
- A recurring source segment within the analysed files. Its count can depend on file order and analysis settings. Identical wording may still require a context check before one target form is reused throughout a product.
- CAT match analysis
- A report grouping content by repetitions and similarity to available memory data. It supports scoping and pricing, but its meaning depends on selected memories, thresholds, exclusions and the exact source package analysed.
- Weighted word count
- A commercial calculation applying agreed weights to analysis categories. It is distinct from the actual source volume. A buyer needs both the underlying counts and the weighting rules to understand the resulting chargeable quantity.
- Machine translation (MT)
- Automatic generation of target-language text by a computational system. Its output can contain fluent errors, so intended use, processing permission and the human checking process need to be agreed before it enters a project.
- Machine translation post-editing (MTPE)
- Human correction of machine-generated translation against an agreed quality target. The scope should specify the content, checking criteria and deliverables; the label alone does not establish how thoroughly the output will be assessed.
- Light post-editing
- Post-editing to an explicitly limited specification, often prioritising understandable meaning over publication-ready style. The limitations need to be written down. It should not be presented as the full post-editing process specified by ISO 18587:2017.
- Full post-editing
- Post-editing intended to produce a complete, accurate and linguistically suitable target text under agreed specifications. ISO 18587:2017 addresses full human post-editing and post-editor competence; it does not make every machine-generated draft suitable for that workflow.
- Translator’s check
- The translator’s examination of the completed translation before it moves to the next stage. This self-check can correct omissions and inconsistencies, but it remains distinct from a second person’s bilingual revision.
- Revision
- Bilingual examination of a translation against its source, assessing accuracy and suitability for the agreed purpose. In ISO 17100 it is performed by someone other than the translator; target-only reading cannot serve the same function.
- Review
- Examination of target text for suitability to its intended purpose and domain conventions. Its exact scope should be specified. Specialist review can complement bilingual revision, especially when a product owner needs to approve terminology or technical presentation.
- Proofreading
- Checking proofs before publication, with attention to the final text and presentation. It should not be used as an undefined synonym for translation revision; the supplied material and checking responsibilities need to be stated.
- Back translation
- Translating a target text back into the original language, usually for a defined evaluation purpose. Differences can reveal issues but do not automatically prove an error; comparison needs context and a documented assessment method.
- Linguistic quality assurance (LQA)
- Assessment of language quality against defined criteria, sometimes using error categories and severity levels. A meaningful LQA specification states the sample or coverage, evaluator role, scoring method and how disagreements are resolved.
- Automated QA
- Software checks for detectable patterns such as inconsistent numbers, missing tags or terminology mismatches. Findings require interpretation, and a clean report does not demonstrate that meaning, grammar and domain usage are all correct.
- DTP
- Desktop publishing: preparing text and visual elements in a page layout. Multilingual DTP can involve changed line lengths, fonts and page flow. The term describes a production activity, not an implied Babylon service.
- Placeholder
- A token replaced by a value at runtime, such as a customer name or quantity. Translation instructions should explain its meaning and syntax, including whether reordering is permitted and which characters must remain unchanged.
- Inline tag
- A marker inside text representing formatting or another structural element. Tags often need to be preserved and correctly positioned. Their visible notation varies by tool, so editing instructions should identify protected elements explicitly.
- String ID
- A stable identifier connecting a localisable text unit to an application or content system. It helps track updates and queries even when source wording changes, and should remain distinct from the translated display text.
- Pseudolocalisation
- A technical test using modified text to expose layout, encoding or extraction problems before real translation. It can reveal fixed widths and hard-coded strings, but it does not assess the quality of a Bulgarian translation.
- Text expansion
- An increase in visible text length after translation or adaptation. It varies by phrase and context. Layout testing with actual content is more reliable than applying one expansion percentage to every screen.
- Source freeze
- An agreed point at which a particular source version becomes the production baseline. Later amendments remain possible, but need a change record and an assessment of affected files, languages, costs and approval milestones.
- Terminology query
- A recorded question about the intended meaning or preferred equivalent of a term. A useful query includes context and a proposed interpretation, allowing an authorised owner to make a decision that can be reused consistently.
- Standard page
- A defined volume unit rather than a physical sheet. 1,800 characters including spaces is a common convention in several Central and Eastern European markets, while other markets count words. The agreement should identify source or target counting and the rounding method.
- Sworn translation
- A term for translation associated with an authorised translator under a jurisdiction-specific system. Its meaning depends on the receiving country and institution. This definition does not state that Babylon provides sworn translation or any related formalities.
- Interpreting
- Rendering spoken or signed communication into another language. It is distinct from written translation, and its planning needs include the event format, language directions, speakers, participants and the appropriate interpreting mode.
- Simultaneous interpreting
- Interpreting while the speaker continues, with a short processing delay. The event specification needs to identify the working conditions and delivery arrangement; the mode itself does not establish which equipment or platform a provider supplies.
- Consecutive interpreting
- Interpreting after the speaker pauses or completes a section of speech. The meeting schedule needs space for the interpreted contribution, and speakers should agree how turns and pauses will be managed.
- Whispered interpreting
- A form of simultaneous interpreting delivered quietly to a very small group near the interpreter. Its suitability depends on acoustics and participants; it is a distinct arrangement that should be agreed for the specific event.
- Liaison interpreting
- Interpreting dialogue between participants, commonly through short consecutive exchanges. The term often describes a meeting setting rather than a wholly separate technical mode. Language direction, turn-taking and the interpreter’s role need clear agreement.
Turn terminology into a usable brief
The definitions become useful when a buyer links them to deliverables: translation plus revision, an approved termbase, an agreed file export and named acceptance criteria. The B2B preparation guide provides that handoff structure. A specialist term should clarify responsibility, rather than make a quotation harder to understand.
For event planning, the interpreting page separates spoken-language work from document translation. For digital releases, locale behaviour and file handling need their own specification. A provider’s use of a term does not imply that every associated activity is included in the same price or supported for every project.
Buyer questions
Practical
answers
Is a CAT tool the same as machine translation?
A CAT tool and machine translation are different concepts. A CAT environment helps a linguist handle segments, terminology, memory suggestions and formatting. Machine translation generates proposed target text automatically. A CAT tool may connect to an MT system, but it can also support a human translation workflow without using machine-generated output.
What is the difference between a glossary and a translation memory?
A glossary records terms and their meanings or approved equivalents, while a translation memory stores paired source and target segments. Both need an approval and maintenance process. A memory example can provide context, but it should not automatically override a current approved term where the historical translation uses obsolete wording.
Does a 100% match mean the translation is correct?
A 100% match means that the source matches a stored source segment under the tool’s rules. It does not independently validate the target translation, its approval status or its suitability for a new context. A buyer should ask how matches are checked and whether analysis settings or memory quality affect the proposed workflow.
Is light post-editing covered by ISO 18587?
ISO 18587:2017 addresses full human post-editing, rather than an undefined light-editing service. A limited post-editing assignment therefore needs its own clear quality specification and intended use. Labels such as light and full should be connected to actual acceptance criteria. The ISO catalogue’s scope was checked 31 October 2026.
Are revision and proofreading interchangeable?
Revision and proofreading describe different checks. Revision examines a translation against its source, while proofreading checks proofs before publication. A specification should identify the material, coverage and purpose of each stage. The word proofreading alone does not establish that a qualified second person has checked the full translation bilingually.
References
Sources and further reading
- ISO 17100:2015 — Translation servicesInternational Organization for Standardization
- ISO 18587:2017 — Full human post-editingInternational Organization for Standardization
- XLIFF Version 2.1OASIS
- TermBase eXchange (TBX) overviewTBX information project
- Translation Memory eXchange (TMX)DeepL Help Centre
- Localisation and internationalisationWorld Wide Web Consortium (W3C)
- Performing translation-memory lookupsRWS developer documentation
- Types of interpretingChartered Institute of Linguists
Sources checked 31 October 2026. Standards summaries explain scope; published standards and applicable legislation govern individual projects.