Text & LLM data · Native reviewers

Translation and localization quality rating data

App screens, product pages, help articles and marketing copy, translated by machine or by people, rated in context by native speakers of each market. Every error is marked by type and severity with a corrected version, so you know which translations read as if they were written there.

Specification

Markets
Mexican and European Spanish, French, German and Japanese, with other languages and locales
Content
App and website UI text, product pages, help articles, marketing copy and chatbot replies
Context
Text is rated on the screen or page where it appears, so length, grammatical gender and placeholders are checked too
Ratings
Accuracy, fluency, terminology, style and tone, local conventions (dates, currency, formality) and an overall score
Annotation
Error spans with type and severity in an MQM-style scheme, and a corrected translation
Quality check
Several native reviewers per segment; a senior reviewer settles any disagreement

Example of the data

What you see, and the file you get.

Preview
es-MXSpain-only word “ordenador”locale · minor
de-DEinformal “du” in a banking appformality · major
ja-JPbutton text cut offlayout · minor
segments.csvCSV · JSON · Excel
segment_idlocalesource_entranslationerror_typeseveritycorrection
ui_0001es-MXSave to your computerGuardar en tu ordenadorlocale_termminorGuardar en tu computadora
ui_0002de-DECheck your balancePrüf deinen KontostandformalitymajorPrüfen Sie Ihren Kontostand
ui_0003fr-FRFree shipping on orders over €50Livraison gratuite dès 50 € d’achat
ui_0004ja-JPAdd to cartカートに追加するstyleminorカートに追加
ui_0005es-ESYour order has shippedTu orden fue enviadalocale_termminorTu pedido se ha enviado

Sample rows. Also as JSON or Excel.

Contact us to get the dataset

Tell us what you need. We walk you through the data, its formats and how it is checked on a short call.

Contact us

What it is used for

  • Choosing a translation engine

    Compare machine translation and LLM output, language by language, on your kind of content.

  • Training quality estimation

    Teach a model to flag bad translations before a person has to.

  • Localization QA before release

    Catch the wrong formality, words from the wrong country and broken UI text before users do.

  • Fine-tuning translation models

    Use the corrected translations as examples of how each market really writes.

Questions

Why rate translations in context?

Text can be right on its own and wrong on the screen: too long for the button, the wrong gender for the item, or too casual for the product. Reviewers see what users see.

Which error types are marked?

Accuracy, fluency, terminology, style, local conventions and layout problems, each with a severity, in a scheme close to MQM.

Can I see samples first?

Yes. Request samples with the form on this page and we will send them with the full specification.

Tell us what your model needs.

Request samples and the full specification. We reply within a couple of business days.

Request samples