Language teams
Which issues recur in this sample, and which terms need a shared decision?
Start with your content requirements and glossary.
A scoped human review of English/Greek interfaces, support content or AI output. Reviewer fit, availability and delivery terms are agreed before work begins.
Source phrase, Englishand a pencil
Suggested revision, Greekκαι ένα μολύβι
Compare a source sentence with a candidate version, then read the reason for a suggested change.
Example 1: Missing meaning.
Please bring a notebook and a pencil to the workshop.
Παρακαλώ φέρτε ένα σημειωματάριο στο εργαστήριο. 1
Παρακαλώ φέρτε ένα σημειωματάριο και ένα μολύβι στο εργαστήριο.
Illustrative example. Original sample text, not client data or a live evaluation.
The examples show how a note can explain a judgment. Labels describe these samples; they are not calculated scores or a language-coverage promise.
Illustrative English/Greek UI release extract. Authored example; no customer result or universal quality score is claimed.
The proposal should make these choices explicit.
Name the release, comparison or content decision the review should inform. Choose the language pairs and domain.
Agree which material will be examined, what counts as an issue and how its importance will be judged.
Specify the language and subject knowledge needed, calibration approach and how disagreements would be resolved. Confirm coverage.
Set the reporting format, annotated examples, any scoring rules and the limits of what the sample can support.
Material sharing, data handling, timing and availability belong in the proposal. Do not include confidential source material in the first enquiry.
Which issues recur in this sample, and which terms need a shared decision?
Start with your content requirements and glossary.
How should a chosen language pair be reviewed after a prompt or model change?
Define the product context and comparison first.
What sample, rubric and reporting would make this comparison useful?
State the assumptions and limits alongside the results.
A review note should identify the relevant words, the requirement being applied and the reason for the judgment.
Agree error categories and scoring rules for the task rather than treating one overall score as the whole report.
Method references
MQM is a framework for structured translation-error analysis.
The WMT 2024 metrics task is an example of comparing metrics with human judgments.
These are method references, not Centaword credentials.
Tell us the language pairs and domain. Reviewer requirements and coverage need to be confirmed in the pilot proposal.
Start with a non-confidential description. Agree material sharing and data handling separately.
It prepares an editable draft and sends nothing. Current email availability and the available copy/open actions are shown beside the form.
Describe the system, languages and question you want to investigate. Prepare an enquiry to discuss the sample, reviewer coverage and reporting.
Or write to
hello@centaword.com
Preparing an enquiry keeps your details in this page. You can edit and copy the draft. How this website handles your information.