Independent Translation Audit
Source-to-target evaluation of existing translations, with categorised findings and an agreed rating approach.
PangeaVox · Independent Language Quality Evaluation
Linguistic Quality Assurance (LQA)
Know how good your translations really are.
Measure the quality of existing human, machine-translated and AI-assisted content with independent linguistic evaluation, documented error classifications and actionable findings.
From a one-off translation audit to a repeatable multilingual vendor assessment, we help you make evidence-based decisions about content quality.
Independent audits · MQM-based error analysis · Human, MT & AI output
LQA reviewIllustrative
SourceSave changes
Translation under reviewSupprimer les modifications
Finding
The translation describes deleting rather than saving changes. Meaning is reversed.
Suggested correctionEnregistrer les modifications
Example only. No client data or actual evaluation score is shown.
What LQA is
Independent Evaluation, Not Just Another Proofread
Linguistic Quality Assurance evaluates a translated text against its source, intended audience and agreed requirements. Errors are identified, categorised and assigned a severity level where relevant.
Unlike editing or proofreading, an LQA audit primarily establishes and documents the quality of an existing translation. Corrective editing may be commissioned separately.
Use LQA when you need to…
- Validate translations before release
- Assess external language suppliers
- Compare human, MT and AI-assisted output
- Investigate recurring terminology or accuracy issues
- Establish measurable, repeatable quality criteria
Choose an evaluation scope
LQA Services for Every Content Workflow
Start with a representative sample or request an extensive evaluation, depending on the volume, risk and intended use of your content.
MT & AI Output Evaluation
Human review of raw machine translation, post-edited MT or AI-assisted output against the same quality criteria.
Vendor Quality Assessment
Comparable evaluations of translations from different vendors or teams, using a consistent specification and sample plan.
Software & Website LQA
In-context assessment of localised interfaces, strings, navigation, terminology and linguistic fit, where test access is available.
Specialist Content Audit
Subject-matter review for technical, medical, legal, financial and regulated content, with the appropriate level of expertise.
Recurring Quality Monitoring
Periodic sampling and reporting to track quality changes, recurrent issues and corrective priorities over time.
A transparent evaluation framework
ISO 5060-Informed Evaluation, MQM-Based Classification
Our proposed evaluation methodology draws on the principles of ISO 5060:2024, which provides guidance on evaluating translation output, including human translation, post-edited MT and unedited MT.
For structured error annotation, we can use an agreed subset of Multidimensional Quality Metrics (MQM), adapted to your content type and quality requirements.
The evaluation brief defines error categories, severity rules, sampling and any weighting or acceptance threshold before the review starts.
Documented, not subjective
What Our Evaluators Look For
The precise taxonomy is agreed for each project. Common MQM-aligned categories include:
Accuracy
Mistranslation, omissions, additions and changes in meaning.
Terminology
Incorrect or inconsistent use of approved terminology.
Linguistic Conventions
Grammar, spelling, punctuation and language conventions.
Style
Tone, register, clarity and compliance with an agreed style guide.
Locale Conventions
Dates, numbers, currency, measurement units and locale-specific formats.
Audience Appropriateness
Wording or references unsuitable for the intended audience.
Design & Markup
Text presentation, tags, formatting and relevant layout constraints.
Impact-based assessment
Severity Levels That Reflect Business Risk
Our evaluators consider how each issue affects meaning, usability and the intended purpose of the translation. Severity definitions are calibrated to the content and agreed with the client.
A numerical score or pass/fail result is provided only if the scoring formula and threshold have been agreed in advance. No universal pass mark is assumed.
Neutral
Acceptable variation or an observation that carries no penalty.
Minor
A limited issue that does not seriously undermine intended use.
Major
An issue that seriously affects accuracy, clarity or usability.
Critical
An issue that makes content unfit for purpose or creates serious risk.
Evidence you can act on
What You Receive
We tailor reporting to your procurement, publishing or vendor-management needs.
Segment-Level Error Log
Original segment identifiers, source and target excerpts, error category, severity and evidence-based explanations. Suggested corrections can be included in the agreed scope.
Executive Quality Summary
Scope, sampling assumptions, issue patterns, risk observations and recommended next steps, with a score or rating if agreed.
Comparable Vendor Results
Consistent evaluation criteria for comparing batches, linguists or suppliers, with a clear explanation of comparability limitations.
Review & Arbitration Option
Optional documented resolution of contested findings and an agreed final audit record.
How it works
From Evaluation Brief to Actionable Report
- 01
Define the brief
Confirm languages, intended use, content sensitivity, methodology, evaluator expertise and any pass criteria.
- 02
Agree the sample
Decide whether to assess the complete content or an explicitly defined sample; record selection and limitations.
- 03
Evaluate independently
A qualified evaluator reviews source and target content and logs substantiated issues by type and severity.
- 04
Consolidate findings
Check annotations for internal consistency and prepare a summary; calculate a score only if agreed.
- 05
Deliver & follow up
Provide the error log and report. Separate remediation, re-evaluation or recurring audits can be arranged.
Frequently asked questions
Linguistic Quality Assurance FAQs
How is LQA different from proofreading or editing?
LQA evaluates and documents the quality of an existing translation against an agreed framework. Proofreading and editing primarily aim to improve or correct the text. Corrective work can be quoted separately after an LQA audit.
Can PangeaVox evaluate translations produced by another supplier?
Yes. We can arrange independent, vendor-neutral evaluations using a common assessment brief. Any potential conflicts of interest are considered when selecting reviewers.
Do you assess AI-generated and machine-translated content?
Yes. Human evaluation can be applied to raw MT, post-edited MT and AI-assisted translation, with categories and severity definitions adapted to the content and risks.
Do you use ISO 5060 and MQM?
Our evaluation approach can be informed by ISO 5060:2024 and use an agreed MQM-based error taxonomy. The exact methodology, scoring formula and deliverables are specified for each project. References to these frameworks are not claims of certification.
Will the report provide a quality score or pass/fail decision?
If required, yes, provided the sampling plan, severity weights, score formula and pass threshold are agreed before evaluation. A sample cannot guarantee that an unreviewed translation is error-free.
Can you correct every error found?
Corrective editing, post-editing or full revision can be ordered separately. An LQA audit itself focuses on objective evaluation and reporting.
What files do you need for a quotation?
Ideally, provide the source and target texts, language pairs, approximate word count, content type, intended use, deadline and any reference materials or quality requirements. Please arrange secure transfer before sharing confidential or sensitive files.
Independent evaluation. Clear findings.
Need an Independent Translation Quality Assessment?
Tell us which languages, content types and translation outputs you need evaluated. We will recommend a suitable sampling plan, assessment framework and reporting format.