Welcome to the ATA Corpus, a unique and searchable body of translated texts, created by the BYU Translation Research Group with the support of professors from language and computer science departments. This resource is designed for educators, researchers, and anyone interested in translation studies.

How to use it?

The ATA Corpus is not just a database—it's a curated corpus of translation exam texts gathered in partnership with the American Translators Association (ATA). What sets this corpus apart is its depth of linguistic annotation: each translation error is marked with both its type and severity. These annotated texts serve as a valuable resource for language instructors, translation trainers, linguists, and for advancing research in translation error analysis and machine translation. Featuring robust exam reconstruction and search tools, the ATA Corpus allows users to explore hundreds of bilingual exam passages across multiple languages, filter and examine different error types, and reveal translation practices over the years.

ATA Corpus Help Page

Below you'll find answers and overviews for core features and concepts within the ATA Corpus platform.

What is an ATA Exam?

The ATA certification exam is a rigorous test in which candidates translate a set of passages from a source language into a target language. Each candidate receives:

  • Three passages:
    • Passage A: General content (must be completed to pass)
    • Passage B: Scientific or medical content
    • Passage C: Legal or business content
  • Candidates must translate Passage A plus either Passage B or C.

Exams are handwritten and subsequently anonymized to protect test takers' and graders' identities. The ATA provides all exam data and mandates the removal of any identifying marks. Grading is performed by ATA-certified reviewers, nearly all of whom are highly proficient native speakers of the target language. Each completed translation is scored using the ATA error framework; every error is logged with a specific type and severity, contributing to the candidate's overall score. Errors are also cross-mapped to the MQM (Multidimensional Quality Metrics) standard for advanced research and comparison purposes.

Generally, each exam is reviewed by two graders, with exceptions for incomplete exams or significant error thresholds. If graders disagree, a third grader may be assigned. Candidates must not exceed 18 error points to pass.

Saved Queries