Public documents are read by people who have no choice about reading them
A citizen cannot go to a competitor for a clearer version of a tax form, an asylum decision or a public health notice. That asymmetry is why the standard for public-sector language should be higher than for commercial content, not lower.
What we work on
- Tender and procurement documentation
- Statutory and regulatory publications
- Citizen-facing information and forms
- Public health and safety communication
- Immigration, asylum and social services documentation
- Policy papers and consultation material
- Institutional and inter-governmental content
- Public sector training and internal guidance
Why this content deserves independent review
The reader has no alternative and often no leverage. Commercial content that reads badly loses a customer. Public content that reads badly leaves someone unable to claim a benefit, meet a deadline, understand a diagnosis, or respond to a legal decision that affects their status. The consequence lands on the person least equipped to absorb it.
Comprehensibility is the requirement, not just accuracy. Public information is written for a general readership, frequently including people reading in a second language or with limited formal education. Machine output preserves the register of the source, and the source is usually institutional. Accurate and impenetrable is a failed public document.
Language variety is a real decision in migrant-facing content. Information translated into a formal standard variety may be unreadable to the community it is aimed at. In Arabic this is acute: the difference between Modern Standard Arabic and the regional varieties is large, and an engine defaults to the standard when uncertain. In the Indic and African language sets the same problem appears with script and dialect choices.
Procurement documentation is legally consequential. Tender text defines what bidders are committing to. Ambiguity introduced in translation becomes a challenge, a delay, or a dispute.
Everything is published and permanent. Public documents are archived, cited, compared across versions and read by people looking for inconsistencies.
What we deliver
- Independent review by reviewers matched to the subject — legal, health, social policy — with documented qualifications available on request.
- A comprehensibility assessment as a named category, separate from accuracy, for citizen-facing text.
- An explicit variety and register decision for each target audience, recorded rather than defaulted.
- Structured error reporting against MQM, with anything that would change a citizen's understanding of a right, an obligation or a deadline treated as a critical finding.
- A version trail suitable for publication records and audit.
- Terminology held across a programme, so the form, the guidance notes and the decision letter use the same words.
Languages where this matters most
Arabic, for migrant-facing and Gulf-facing content, and because it is the language where machine translation performs worst on the best independent evidence available. The Indic and African language sets, where the training data behind the engines is thin regardless of how many people speak the language. Ukrainian, Somali, Tigrinya, Pashto, Dari and the other languages that appear in resettlement and asylum work — where supply of qualified reviewers is genuinely scarce and where a bad translation is not a customer-service problem.
The full EU official set for institutional content, where the requirement is every language and the exposure is whichever one nobody in the department can read.