Multidimensional Quality Metrics
Feedback on these documents is welcome and comments can be sent to feedback [at] qt21.eu
Last updated: 2014 November 21
What is MQM?
MQM stands for Multidimensional Quality Metrics. MQM is a framework for describing translation quality metrics in a consistent and coherent fashion and tying them to specific project requirements. With MQM you can check the translation quality issues that matter to you, in as much or as little detail as is needed. MQM supports both quick, high-level, “holistic” evaluation that focuses on entire texts and detailed “analytic” evaluation that identifies specific problems in text. It is easily adaptable to manual, semi-automatic, and automatic evaluation and can be applied to any text type. It allows quality to be evaluated along multiple dimensions, allowing you to identify specific problems and understand the strengthes and weaknesses of specific translations.
MQM is not a one-size-fits all translation metric. It does not tell you to check a lot of things that do not matter for a particular task, but instead focuses on what matters for each task. It does not provide just a single score, but instead provides insight into the specific qualities of each translation.
Because MQM is based on existing best practices, it can be easily implemented in a variety of environments (from spreadsheets to dedicated evaluation environments) and does not require you to overhaul your existing quality practices while providing an easy path to improve them.
MQM’s central features include the following:
- At the heart of MQM is a descriptive listing of over 100 types of translation quality issues derived from a careful study of existing translation quality evaluation metrics. These types are arranged in a hierarchy, and range from broad to detailed, allowing you to obtain as much or as little detail as you need. These issues cover a wide variety of issues, from purely translation issues (such as Mistranslation) to design, and from problems with the target language to suitability of the translated text for the target market. Because these issues are clearly defined, evaluators are able to provide clear and objective insights into the specifics of translation quality.
- MQM is based on standards. It utilizes translation specifications (based on the ASTM F2575 for describing translation project parameters) to define translation requirements. It interfaces well with the localization quality issue data category in the W3C’s Internationalization Tag Set 2.0 specification. It provides a XML-based format for declaring metrics and marking issues in translations.
- MQM is multi-dimensional. It allows users to select issues that measure translation quality in multiple dimensions—Accuracy, Fluency, Design (ie., layout and formatting), Verity (a new way of dealing with the suitability of text for the target locale and audience), and Internationalization (issues related to whether the source content was properly prepared for translation)—and assess quality for each dimension. Thus, rather than providing a blanket score, MQM provides insight into the specific strengths and weaknesses of particular translations.
- Sign up for the QTLaunchPad mailing list by sending an email to info [at] qt21.eu
- Join the QTLaunchPad LinkedIn group.
- Visit the QTLaunchPad Facebook page
The following links will help you learn more about the QTLaunchPad Multidimensional Quality Metrics:
- Background and Design Principles
- Technical Details