A Methodology for Generating Items in Three or More Languages Using Automated Processes
Educational and psychological tests are administered to examinees in different languages across different cultures throughout the world. The challenges inherent to translating and adapting multilingual and multicultural assessment are enormous. The purpose of this paper is to describe and illustrate a new methodology that can be used to generate items in multiple languages. The method is presented as a three-stage process where, first, context translation begins when the context of the model required for item generation is translated or adapted appropriately for each language group; second, words and key phrases are translated; third, content assembly occurs where computer algorithms place the words and key phrases into the context-specific item model. Then, we demonstrate how the method can be applied to a diverse sample of item models in math and science to generate thousands of multilingual test items. Finally, results are presented from a substantive review designed to evaluate item quality which revealed that 91% of the generated items were judged to be acceptable by two bilingual test development specialists. Directions for future research are also presented.