Type of Project: Funded project Joint project
Including professional exams without language barriers through text optimization with the help of artificial intelligence (TOP.Kl)

People with impairments have poorer prospects on the labor market, not least because increasingly complex work processes require ever higher qualifications. Examinations associated with this determine life paths. For people with limitations or disabilities, they can mark their entry into a working life with employment subject to social security contributions on the primary labor market - or a failure of years of effort, often combined with individual support. A major challenge is that examination tasks and exercises are often formulated in a way that is very difficult to understand. The language level of many learning materials - due to their complexity - exceeds the abilities of many trainees. This means, among other things, that the ability to read and write becomes a selection criterion both at school and in a job, regardless of whether the realities of work actually require these skills later on. At the same time, examinations do not lose any of their technical quality and requirement level if they are formulated in a way that is easy to understand. Making textual adaptations for individual tasks is time-consuming and currently has to be done by experts. Unfortunately, there are currently no ready-to-use translation systems known that are generally available for the German language on the one hand and that are specifically optimized for the required range of teaching and examination text phenomena on the other. In the TOP.KI project, an AI translation tool for examination tasks in simple language is being developed, tested and evaluated. A service will be created that (similar to the translation tool DeepL) receives text in formulated form, processes it and returns it in Simple Language. A wide variety of aspects and problems in the field of educational language are included at text, sentence and word level. Based on the tool, significant experience and knowledge can be gained to adapt further job learning content to the needs of people with impairments. To this end, the AI tool will be made widely available. The focus on people with impairments is also to be dissolved in a further step. In cooperation with training and testing institutions (especially IHK and HWK), proposals are being drafted to make the exams text-optimized with AI available to as large a group of trainees as possible. It is to be examined whether the text optimization can even benefit all trainees, so that there is no longer any need for examination differentiation. The KI tool for occupation-specific technical language should help to make the linguistic requirements in examinations visible and controllable even for linguistic laypersons and in this way reduce language-related barriers in examinations. This would help to create linguistic equality in examinations, because standard language barriers have a negative effect on technical performance and lengthen the processing time. Project partners are the Institute for Text Optimization, Halle/Saale, the Central Technical Committee for Print and Media, Kassel, the University of Wuppertal, the vocational training centers Oberlinhaus Potsdam and Leipzig, and deepsight GmbH in Osnabrück.