Skip to toolbar

Community & Business Groups

MQM Terminology

This document is a working list of terms for various aspects of MQM.

TermPart of speechDefinitionDiscussion
Quality management
quality planningn—quality management activities for designing a system of policies, processes, and procedures capable of producing deliverables that will fulfill stakeholder requirements
quality assurance (QA)n—quality management activities that have the objective of auditing processes and procedures to provide confidence that stakeholder requirements can be fulfilledDiscussion: Quality assurance is not equivalent to quality control.
quality control (QC)n—quality management activities for monitoring and assessing real-time performance in order to verify that stakeholder requirements are being fulfilled within prescribed limits
quality improvementn—quality management activities focused on preventing variation from stakeholder requirements by adjusting a process—including any changes to measurements, resources, methods, tools, and training—to increase its ability to produce quality deliverables
quality evaluation (QE)n—quality management activities for determining whether stakeholder requirements have been fulfilled through inspection and measurement of product propertiesDiscussion: Quality assessment is not a valid synonym for quality assurance or for quality evaluation with respect to this standard, and the initialism QA refers to quality assurance.
Translation, translation specifications
translation (T9n)n—set of processes to render textual content from one language into an equivalent written or digital form in another languageDiscussion: In general usage, the word translation can refer to the translation process or to a translation product.
sourceadj—(term specific to this standard) of or pertaining to the language or culture in which content to be translated was created.
 
Alternative: of or pertaining to the environment for which the original text was created
Discussion: When used as a n, synonymous with source content or source text.
targetadj—(term specific to this standard) of or pertaining to the language or culture into which translated content will be consumed
 
Alterative: of or pertaining to the environment for which a translation product is created
Discussion: When used as a n, synonymous with target content or target text.
segmentn—text string that is an operational translation unit in a computer-assisted translation tool and the basis for translation equivalences in a translation memoryDiscussion: Translation segments typically comprise sentences or sentence fragments, such as a headings, list items, or labels, but can in some instances include entire paragraphs.
translation memoryn—database of source-language segments, usually sentences or sentence fragments, aligned with previously translated target-language segments created in a computer-assisted translation tool or by an automatic segment alignment tool
 
Alternative from F 2575: text-based resource consisting of aligned text segments (translation units) stored by a translation memory tool.
translation modalityn—means by which a translation product is created, with respect to human and machine translation processesDiscussion: At the highest level, translation modalities comprise human translation, unedited machine translation, and post-edited machine translation.
contentn—information with a purpose in the form of text, video, audio, photos, or other media
textn—content in written form
translation productn—translated content as formatted and laid out in a document, web page, or application user interface, including text and complementary components, such as graphics, video, hyperlinks, and accessibility content
requirementn—need or expectation that is stated, generally implied, or obligatoryDiscussion: Typical specified translation requirements can include using a specific termbase or style guide or targeting a specific audience
specificationn—document that sets out detailed requirements to be satisfied by a translation product, and the procedures for checking conformity to these requirements (Based on: ISO 6707-2:2017(en), 3.2.22)
style guiden—specifications designed to ensure consistency and correctness of [information for use] […of a writing or translation product]
Based on ISO …[Tech Comm WG vocabulary standard]
 
normative specification of general linguistic, stylistic, and formatting guidelines for writing, translating, and editing text, with normative term glossaries
Error, error type, error typology
errorn—violation of a rule of good writing or good translation according to specifications
error attributen—defining component of an error assigned in error annotation and referenced in error compilation and quality analysisDiscussion: In MQM, the error attributes are the three defining characteristics of an error—error type, error severity level, and error root cause.
issuen—(term specific to this standard) potential error, typically identified automatically, that has not yet been examined by a human quality evaluator to determine whether it is a valid error or a false positive
potential errorn—
error typen—class of errors identified by error type IDs, error type names, definitions, and positions in a hierarchical organization
error categoryn—
error typologyn— taxonomy of error types, including their error types, that can be assigned to errors in quality evaluations in order to characterize the nature of problems encountered
MQM error typologyn—analytic writing and translation error typology underlying the MQM quality evaluation model, organized under seven/eight top-level error type dimensions
DQF-MQM error typologyn—subset of the MQM error typology developed by the Translation Automation User Society (TAUS) to support software localization as the quality metric component of DQF
MQM-compliant error typology,
compliant error typology
n—subset of the MQM error typology, optionally extended with valid user-defined error types
error annotationn—review by a human quality evaluator to find and flag errors manually in a writing product or translation product
manual error annotation,
human error annotation
n—standard mode of an error annotation tool in which human evaluators find and flag errors in all active annotation error types
automated error annotationn—mode of an error annotation tool in which a system process automatically finds and flags issues corresponding to active annotation error types
recurring errorn—one of two or more errors that share the same error string and the same error type, for an error type that supports recurring error conflation
recurring error conflationprocessing option in error compilation, for certain error types, that allows multiple instances of those error types in the error list that share the same error string to be assigned a reduced error count in the error summary
error severity leveln—one of a small set of error severity designations, ranging from neutral to critical, reflecting the effect of the error on the usability of the text
quality scoren—quality measure of a writing product or translation product, a multiple, usually 100, of the difference between 1 and the normed penalty totalDiscussion: The maximum and best value for a quality score is the maximum quality score, usually 100.
error root causen—proximate cause at the end of a causal chain that leads to and is responsible for an error in the evaluation text
Metric
metric n—standard of measurement that defines the conditions and the rules for performing a measurement and for understanding the results of a measurement (ISO/IEC 19086-2:2018(en), Cloud computing — Service level agreement (SLA) framework — Part 2: Metric model, 3.6)
MQM-compliant metric,
compliant metric
n—translation quality metric based on an MQM-compliant error typology and an MQM-compliant scoring model
Quality analysis/evaluation/assessment
quality analysisn—application of scoring parameters and a scoring model to the error counts output in an error annotation project to automatically calculate quality measures and quality ratingsReview
full quality analysisquality evaluation project in which quality measures generated in quality analysis are based on the error counts of the full set of active annotation error types or the full set of active root causesDiscussion: Normed penalty totals and quality scores based on a full quality analysis are the default, unmarked case. They are only specified as FQA-based to contrast them to PQA-based quality measures.
partial quality analysisn—quality evaluation project in which quality measures generated in quality analysis are based on the error counts of a proper subset of the active annotation error types or a proper subset of active root causesDiscussion: Without the application of a partial-analysis scalar, quality measures would improve artificially in a quality evaluation with partial-analysis error counts and, as a result, inflate and distort the quality measures generated.
translation quality evaluation (TQE)n—quality evaluation of a translation productDiscussion 1: The terms “translation quality assessment” and “TQA” used in the sense of translation quality evaluation is deprecated in the context of this standard in order to avoid confusion with translation quality assurance, TQA.
analytic translation quality evaluation (analytic TQE),
analytic evaluation
n—quality evaluation that identifies and tallies errors from an analytic metric and calculates quality measures and quality ratings using a suitable scoring modelDiscussion 1: Analytic metric can include the notation of severity levels.
holistic translation quality evaluation,
holistic translation evaluation
n—quality evaluation based on identifying overarching qualities such as readability and accuracy at the macro level
score directionalityn—indication of whether a higher or lower score is better
normed penalty totaln—quality measure of a writing product or translation product that represents the error penalty total relative to the number of words in the evaluation textDiscussion: The minimum and best value for a normed penalty total is 0.
scoring modeln—description of the data points, scoring parameters, and formulas used to calculate quality measures and quality ratings in a quality evaluation
reliabilityn—characteristic of a quality evaluation system that produces stable, reproducible, and consistent results using the same specifications and instructions in repeat evaluations of the same or similar evaluation texts
validityn—characteristic of a metric that effectively measures what it is intended to measure
validation n—
verificationn—process of checking a metric against specifications
 
Alternative: confirmation that project specifications have been fulfilled”
Editor, Reviewer, Evaluator roles
editor,
bilingual editor,
reviser
n—bilingual member of the translation team who compares a completed translation to the source text for the purpose of validating the accuracy of the final target text, and gives detailed feedback [ASTM 2575:2014, 3.1.7]
 
Alternative: person who examines the entire target language content against the source language content to ensure linguistic accuracy and faithfulness to the source language content (Based on ISO 20539, 3.3.9)
reviewer,
monolingual reviewer
n—person who reviews a translation product
 
person who examines the entire translation product in order to ensure its domain accuracy (Based on ISO 20539, 3.3.8)
 
person who examines the target language content in order to ensure its domain accuracy (Based on ISO 20539, 3.3.8)
Discussion: Regardless whether the reviewer is personally mono- or bilingual, the examination at this stage is monolingual with respect to the target language and the domain.
translation quality evaluator,
translation evaluator
n—person qualified as a revisor who conducts a translation quality evaluation of a transation product [produced by another person]person who is qualified as a editor (reviser) who conducts a quality evaluation of a writing product or translation product produced by another human or by a machine translation program
Dimensions
error type dimension,
dimension
n—top-level error type in the MQM error typology
accuracyn—degree to which the target text accurately reflects the intended or assumed meaning of the source text, with nothing added, deleted, or changed, allowing for any differences authorized by translation specifications
terminologyn—set of designations and concepts belonging to one domain or subject [Source: ISO 1087]
fluencyn—linguistic well-formedness of a text segment, including grammatical and mechanical correctness, as well as standard usage Discussion 1: For purposes of this standard, fluency can apply to either the source or the target text.
 
Discussion 2: Fluency focuses on elements that are either objectively correct or incorrect as agreed upon with respect to accepted linguistic norms.
 
Discussion 3: Fluency in this sense should not be confused with mastery of a language in the context of language acquisition.
stylen—appropriateness of a text with respect to general language and any special language requirements with respect to lexical and usage conventions, as well as compliance with the organizational style guide component of the translation specificationsDiscussion: Typical style considerations are monolingual and can include register, brand guidelines, formatting instructions, and compliance with normative references.
designn—physical presentation of a writing product or translation product Discussion: Design includes text, paragraph, and user interface element formatting and markup, integration of text with graphical elements, and overall page or window layout.
locale conventionn—(term specific to this standard) set of locale-specific content or formatting, layout, or other presentational practice anticipated by the intended audience Discussion 1: Locale conventions include such factors as calendar internationalization, number, date , name, address, and telephone formats, as well as shortcut key.
 
Discussion 2: Locale conventions specify local-specific norms for text and data elements.
locale appropriateness (verity)n—(term specific to this standard) suitability of a passage in a translation product with respect to the context and needs of the intended audienceDiscussion 1: Locale appropriateness (verity) factors reflect, for instance, legal requirements, locale-specific content, culture-specific references, end user suitability, completeness, and factual errors.
 
Discussion 2: Locale appropriateness indicates that the inclusion and content of a text passage are correct for the target locale.
extrasegmentaln—features of a text that transcend features of individual text segmentsDiscussion: Extrasegmental features of a text include cohesion, coherence, and intertextuality.
cohesionn—property of a text related to grammatical dependencies in the text that signal relations among the various elements making up the text
coherencen—property of a text whereby the text assumes an associative structure that is logical within the framework of a given language and cultureDiscussion: Coherence functions as an interaction of the pre-knowledge possessed by the reader and the new knowledge presented by the text itself.
intertextualityn—property of a text whereby knowledge presented in the text is closely related to or dependent on knowledge presented in other texts common in a language or culture
Additional terms
quality evaluation stagen—discrete manual or automatic stage in the sequence of processes entailed in an analytic quality evaluationDiscussion: Quality evaluation proceeds, in order, through one manual stage, error annotation, and three automatic stages, error compilation, quality analysis, and quality rating.
error compilationn—organization of data output from an error annotation project in a format that supports quality audits and input to quality analysis processesDiscussion: The two formats supported in error compilation are error lists and error summaries. Error lists have an entry for each error identified in an error annotation project, including essential error identification data, such as error location, error type ID, error string, error severity level, and error root cause. Error summaries tally the errors from the error list to create a matrix of error counts for individual error type–severity level combinations.
quality ratingn—secondary quality category or rank based on the degree to which a writing product or translation product meets requirements for a particular applicationDiscussion: Quality ratings are typically determined by comparing quality measures against a set of related quality grades, but they may be based on any aspect of a writing product, translation product, or quality evaluation.
quality measuren—quantitative measure of a writing product or translation product produced by a quality evaluation process, identified by best score and score directionalityDiscussion: In MQM and other analytic TQE metrics, the two quality measures are normed penalty total and quality score.
termbaseimplementation of a terminology, in particular, a bilingual terminology, that is used and maintained by a terminologist or by a translator working in a translation editor or a standalone terminology management system
 
Alternative: database consisting of concept-oriented terminological entries and related information, usually in multilingual format
root cause taxonomyn—taxonomy of error root causes that can be assigned to errors in quality evaluations to specify the objects or actions that are the proximate sources of the errorsNote 1: A root cause taxonomy is structured as a hierarchy of error root causes, and root causes are identified with root cause IDs, root cause names, definitions, and example root causes.
MQM root cause taxonomyn—root cause taxonomy underlying the MQM quality evaluation model
root cause analysisn—association in error annotation of errors with their root causes, taken from a standard root cause taxonomy, and use of these root causes in error compilation and quality analysis to generate targeted root cause–based quality measuresNote 1: The quality measures produced in quality analysis are generated using partial quality analysis to target specific high-level root causes of interest.
penalty scalarn—value that the normed penalty total is multiplied by to get the quality score into a range that comports with typical ranges of perceived quality, used to calculate quality measures and quality ratings
partial-analysis scalarn—value that, in concert with the penalty scalar, the normed penalty total is multiplied by to calculate quality measures in a quality evaluation that is limited to a partial set of the error counts generated in error annotation
extrasegmentaladj—transcending features of individual text segments, encompassing multiple segments or a whole textDiscussion: Extrasegmental features of a text include cohesion, coherence, and intertextuality.
No Comments |