Research tools for studying how machine translation renders the Turkish evidential opposition between {-DI} (direct) and {-mIş} (indirect) in legally significant statements, and how readers interpret the result.
Compare Claude, GPT, Gemini, DeepL and Google on one sentence under the isolated, combined and instructed conditions, with repeated runs and full logging.
Open the bench → Study 2 · ReadersOpen or close the participant survey, monitor balance across conditions and backgrounds, and export responses.
Manage the experiment → Participant viewGo through the survey exactly as a participant would. Nothing is saved in preview mode.
Preview the survey →