What is MQM?
MQM (Multidimensional Quality Metrics) is a comprehensive framework for assessing translation quality. Developed through collaboration between DFKI (German Research Center for Artificial Intelligence) and the QT21 project, MQM provides a standardized way to identify, categorize, and measure translation errors.
The framework is standardized under ISO 21999 and is widely used in the localization industry for objective quality assessment.
Why MQM Matters
Traditional translation quality assessment often relied on subjective judgments like "good" or "bad." MQM transforms this into a data-driven process by:
- Providing a consistent error taxonomy
- Enabling objective scoring across projects and vendors
- Supporting quality improvement through detailed error analysis
- Facilitating client communication with clear metrics
MQM Error Categories
MQM organizes errors into a hierarchical taxonomy. The main categories include:
Accuracy
Errors related to faithfulness to the source text:
- Mistranslation — incorrect meaning conveyed
- Omission — content from source not translated
- Addition — content added not in source
- Untranslated — source text left as-is
Fluency
Errors affecting readability in the target language:
- Grammar — grammatical errors
- Spelling — orthographic errors
- Typography — formatting issues
- Register — inappropriate formality level
Terminology
Errors in specialized vocabulary:
- Inconsistent — same term translated differently
- Wrong term — incorrect terminology used
Style
Errors related to writing conventions:
- Awkward — unnatural phrasing
- Company style — violations of style guide
Error Severity Levels
Each error is assigned a severity level:
| Severity | Description | Weight |
|---|---|---|
| Critical | Renders content unusable, causes harm | 25+ |
| Major | Significantly impacts meaning or usability | 5-10 |
| Minor | Noticeable but doesn't impact understanding | 1-2 |
MQM Score Calculation
The MQM score is calculated using the formula:
Score = 100 - (Total Penalty Points / Word Count × 100)
Where penalty points are calculated by multiplying error counts by their severity weights.
Example Calculation
For a 1000-word document with:
- 1 critical error (25 points)
- 3 major errors (5 × 3 = 15 points)
- 5 minor errors (1 × 5 = 5 points)
Total penalty: 45 points Score: 100 - (45/1000 × 100) = 95.5
MQM vs Other Frameworks
| Framework | Focus | Standardization |
|---|---|---|
| MQM | Comprehensive error taxonomy | ISO 21999 |
| DQF | Dynamic quality evaluation | TAUS standard |
| LISA QA | QA checklist approach | Legacy (deprecated) |
MQM and DQF are often used together — MQM provides the error taxonomy while DQF adds workflow integration.
Implementing MQM in Your Workflow
1. Define Quality Thresholds
Establish acceptable score ranges for your content types:
- Marketing content: 98+
- Technical documentation: 95+
- Internal communications: 90+
2. Customize Error Weights
Adjust severity weights based on your priorities. For medical content, terminology errors might warrant higher penalties.
3. Train Your Team
Ensure all reviewers understand the error categories and apply them consistently.
4. Use Automated Tools
Platforms like KTTC automate MQM scoring, ensuring consistent evaluation across all projects.
FAQ
What is the difference between MQM and traditional QA?
Traditional QA often uses binary pass/fail criteria or subjective ratings. MQM provides granular error categorization and objective numerical scores, enabling data-driven quality management.
How do I choose which MQM categories to use?
Start with the core categories (Accuracy, Fluency, Terminology) and add domain-specific categories as needed. Most projects don't need the full taxonomy.
What is a good MQM score?
This depends on content type and client requirements. Generally:
- 98+ is excellent
- 95-98 is good for most content
- Below 95 may require revision
Can MQM be used for machine translation evaluation?
Yes, MQM is commonly used to evaluate MT output quality, especially for MTPE (Machine Translation Post-Editing) workflows.
Conclusion
MQM provides the translation industry with a standardized, objective framework for quality assessment. By implementing MQM in your workflow, you can ensure consistent quality, facilitate clear communication with clients, and drive continuous improvement in your translation processes.
KTTC offers built-in MQM scoring, making it easy to implement this framework in your translation quality workflow.