The overall DFG-project aimed to enhance questionnaire translation methodology in cross-national and cross-cultural survey research. Toward this end, we investigated two potential key drivers of questionnaire translation quality, namely, the a) human translators themselves and their backgrounds and b) machine translation, and how this affects survey data quality. In the project submitted here, 32 test persons (16 professional translators and 16 social scientists) produced questionnaire translations from English into German, a first set of items by producing translations from scratch (T1), and a second set of items by producing them based on correcting machine translation output (P2). (The latter process is called post-editing in the translation industry). (Data archived, starting with Exp_). The questionnaires were fielded in a web study, alongside the official published German translation of the items, and a Large Language Model produced translation (ChatGPT, 4.0, free version). This makes altogether 34 different German questionnaire versions of the same source questionnaire. Moreover, the questionnaire was also administered in the English source language for respondents in the UK. The goal was to assess the impact of different translation versions on the resulting data survey quality. (Data archived, starting with S_). This project was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) – project number 502005913.

Nicht-Wahrscheinlichkeitsauswahl - Quotenstichprobe

Selbstausfüller:CAWI(Computerunterstütztes Web-Interview)