Language Data Annotator

Role Overview:

What it is: Review existing translations of English science/knowledge questions into native language Each item has an original translation + an AI-proposed correction. You compare both against the English source and pick the right one. This is a QA/review task, not translation from scratch.

Native Languages Preferred: Bengali, Gujrati, Hindi, Kannada, Malayam, Telugu, Urdu, Odia, Marathi, Nepali, Punjabi, Tamil.

What you'll do:

Check only two things per item: the question and the answer choices.

Assign one of four labels: Accept original, Accept AI correction, Edit final (write your own fix), or Escalate (item can't be safely reviewed).

Judge faithfulness only — same meaning, same number and order of choices, choices stay distinct and grammatically fit the question.

Preserve all numbers, symbols, formulas, units, and technical terms exactly as in the source.

Add a short comment explaining your decision (required whenever you edit).

You must NOT:

Improve style or fluency for its own sake — awkward but accurate is fine.

Fix or reword wrong answers (distractors).

Use your knowledge of the correct answer to change any choice.

Correct mistakes in the original question — copy them faithfully and note it.

You need:

Near-native + strong English reading comprehension.

Comfort reading across topics: sciences, math, computer science, engineering, business, law, humanities.

Sharp eye for grammar (case, gender, agreement, register) and detail.

Discipline to follow strict rules without over-editing.

Bonus: experience in localization QA, data annotation, or linguistics.

You should have experience in any one category mentioned:

• Humanities & language
• Mathematics
• Computer science & IT
• Physical sciences
• School science (elementary)
• Life & environmental sciences
• Engineering & technology
• Business, economics & finance
• Law, governance & philosophy