Technology and Terminology Management
This domain tests working knowledge of translation technology and terminology practice: translation memories, termbases, file formats, quality assurance and machine translation post-editing, including the confidentiality decisions around them.
Translation memory mechanics
Segments are stored as source–target pairs after alignment of legacy material. Leverage depends on segmentation and match bands: an exact match is identical, a context match also matches surrounding segments, and fuzzy matches are graded by similarity. High fuzzy matches still require editing, because a single changed negation or number destroys reuse value. TMX carries translation memories; XLIFF carries bilingual project files; TBX carries terminological data.
Terminology work
A termbase is not a glossary: it stores concept-oriented entries with definitions, subject fields, grammatical information, sources and status labels (preferred, admitted, deprecated). Building one means extracting candidates, verifying them against authoritative sources, and recording the concept - not just the word pair.
Machine translation
Engine types differ, but all output needs post-editing. ISO 18587 distinguishes light post-editing (intelligible, adequate) from full post-editing (publication quality). Confidential material must not be pasted into consumer engines or unvetted cloud services; check the contract and data-processing terms first.
Worked example
A segment returns an 88% fuzzy match reading “payment is due within 30 days”. Reasoning: the similarity score is high but the number and negation are the risk points. Compare the source carefully, confirm the number, then edit and confirm the termbase entry for the payment term. Do not accept on score alone.
Traps that cost marks
- Trusting 100% matches without checking context and placeholders.
- Confusing a termbase with a translation memory.
- Treating post-editing as proofreading rather than rewriting.
- Sending confidential files to a free MT service.