Translation Glossary
Clear explanations of key terms used in the translation industry.
MQM (Multidimensional Quality Metrics)
An international standard framework for measuring translation quality, formalized in ISO 5060:2024. Developed by the EU QTLaunchPad project, MQM classifies translation errors across 7 dimensions (accuracy, terminology, audience appropriateness, design/markup, fluency, locale conventions, style) and severity (critical, major, minor) to produce objective quality scores.
CAT Tool (Computer-Assisted Translation)
Software that assists human translators by providing translation memory matches, terminology lookups, and quality checks. Unlike machine translation, CAT tools augment human expertise rather than replacing it. Examples include SDL Trados, memoQ, and Memsource.
Translation Memory (TM)
A database that stores previously translated source-target segment pairs. When a new text contains segments similar to ones already translated, the TM suggests the previous translation, improving consistency and reducing cost. TMs are stored in TMX format for interoperability.
Termbase (TB)
A structured glossary of approved terminology for a specific project, client, or domain. Unlike dictionaries, termbases prescribe which translations must be used to ensure consistency. Stored in TBX (TermBase eXchange) format per ISO 30042.
XLIFF (XML Localization Interchange File Format)
An XML-based format (OASIS standard) for exchanging localization data between tools. XLIFF 2.0 stores source and target text pairs along with metadata like translation state, notes, and inline formatting. It's the primary data exchange format in the localization industry.
TMX (Translation Memory eXchange)
An XML-based open standard for exchanging translation memory data between CAT tools. TMX files contain source-target segment pairs with metadata such as creation date, translator ID, and language pairs. Defined by LISA (now GALA).
TBX (TermBase eXchange)
An ISO 30042 standard XML format for exchanging terminology data. TBX entries include term, definition, part of speech, context examples, notes, and domain classification. It enables terminology sharing across organizations and tools.
NPT (Normalized Penalty Total)
A metric used in MQM scoring that normalizes total error penalties per 1,000 words of source text. This allows fair comparison across documents of different lengths. Formula: NPT = (Total Penalties ÷ Total Words) × 1,000. Lower NPT = higher quality.
Skopos (Translation Purpose)
A translation theory concept (Vermeer 1978) stating that the purpose of a translation determines the strategies used. A marketing brochure and a legal contract require fundamentally different approaches even in the same language pair.
Back-Translation
The process of translating a translated text back into the source language to verify meaning preservation. Developed by Brislin (1970), it's particularly important in medical, legal, and regulatory translation where meaning accuracy is critical.
LQA (Linguistic Quality Assurance)
A quality assessment process that evaluates translation output against predefined criteria. TAUS's LQA model uses a 4-point adequacy and fluency scale. LQA sampling reviews a representative subset rather than the entire translation.
Post-Editing (PE)
The process of editing machine translation output to achieve acceptable quality. MTPE (Machine Translation Post-Editing) ranges from light PE (fixing only errors) to full PE (achieving human-quality). ISO 18587 defines the standard.
Machine Translation (MT)
Automatic translation by computer without human intervention. Modern MT uses neural networks (NMT) trained on parallel corpora. While fast, raw MT quality varies significantly and typically requires post-editing for professional use.
Terminology Management
The systematic process of identifying, storing, and maintaining terminology for consistent use across translations. Includes term extraction, validation against authoritative sources, approval workflows, and ongoing maintenance.
Segment
A unit of text for translation, typically a sentence or paragraph. Segmentation rules determine how source text is split. Proper segmentation is critical for translation memory matching and consistent translation quality.
Quality Assurance (QA)
Systematic processes to ensure translation meets quality standards. In translation, QA includes terminology checks, formatting verification, numerical accuracy, consistency checks, and final LQA assessment.
Localization (L10n)
Adapting a product or content for a specific locale, going beyond translation to include cultural adaptation, date/number formats, currency, images, and UI layout adjustments. The '10' in L10n represents the 10 letters between L and n.
Fuzzy Match
A partial match between a source segment and a translation memory entry. Typically expressed as a percentage (e.g., 85% match). Higher percentages mean less editing needed. 100% = exact match, 75-99% = fuzzy match, below 75% typically considered new.
Transcreation
Creative translation that adapts marketing or creative content for a target culture while preserving the original's emotional impact and intent. Goes beyond literal translation to recreate the message in a culturally resonant way.