Updated 8 October 2026
Ethics, consent and data
Native speakers give this project their time and judgement. In return we ask before we collect anything, keep only what the research needs, say what we keep, and correct ourselves in public.
Taking part
- Taking part is voluntary, and you can stop at any time without giving a reason.
- Everyone signs an information and consent form first: teachers (18 or older), the head of each school, and, for students under 18, a parent or guardian and then the student. The forms are on the consent page.
- Students take part only if a parent or guardian and the student both sign. Until a partner ethics committee approves the student part, students’ answers are a classroom activity only and are not used in the research.
- Ratings and markings are filed under a code, not a name, and results are published as totals. The sentences being rated come from public datasets and contain no personal information.
- Signed consent forms are kept securely until 31 December 2030 and then deleted. You can ask for your markings to be deleted until 31 December 2030; totals already published in a paper cannot be changed.
What we store
Evaluation platform
What is kept. Your email and password (handled by our database provider's sign-in service), optional background answers (native language, Kashmiri and English proficiency, education level, age group), your ratings, and the time spent on each judgement.
Why. To run blinded comparisons, check agreement between raters, and filter out rushed answers.
Teachers marking translations
What is kept. Your markings, filed under a code such as T01. Your name and contact details appear only on your consent form. Optionally, your age range and home district.
Why. Markings feed research papers and the KashEval benchmark, reported as totals, never by name.
Translator
What is kept. Nothing, when you translate. Only if you answer “Was this translation right?” do we keep the Kashmiri you entered, the translation, your answer and any correction you type.
Why. To find where the model fails and improve it. Don’t include personal information.
Submitting a resource
What is kept. The resource details, and your name, email and note if you choose to give them.
Why. To review the submission and contact you if something is unclear. Submissions are reviewed before they appear.
Browsing the site
What is kept. No advertising or analytics trackers. Your theme choice is saved in your own browser; signing in sets a session cookie. Our host keeps standard server logs.
Why. The site works without accounts; only evaluators sign in.
Data and licences
- Our parallel corpus is built from AI4Bharat’s BPCC and the SMUQamar corpus. The sources are credited, the licence we follow is the one they carry (CC BY-NC-SA 4.0), and the merged data is not redistributed.
- The translator is a research preview for non-commercial use (CC BY-NC 4.0).
- Every photograph and scan on the site is shown with its caption and the credit its licence requires.
- The resource hub links to other people’s work and records each entry’s licence where it is stated. Check the licence at the source before reuse.
Corrections
- When later work changes a story, the post is corrected in place and the correction is noted and dated at the top. The blog shows which posts were updated.
- Results from the ICRTC-2026 paper are reported as published; follow-up work in 2026 is re-testing them, and the site says so wherever they appear.
- Every number we publish is listed with its source on facts and figures. If you find an error, write to us.
Your rights
Under India’s Digital Personal Data Protection Act 2023 you can ask to see, correct or delete your personal data. Write to faizan@kashmirairesearch.online.