AILA - Artificial Intelligence Use in Bilingual Language Assessment

Multilingual children should be examined in all their languages in order to better assess their linguistic abilities. In this way, language development disorders can also be recognised at an early stage because they occur in all languages. In practice, however, there are major challenges in language development diagnostics. One of these is the language barrier: examiners often do not speak the same language as the children and their families.

Artificial intelligence (AI), or more precisely large language models (LLM), can offer a solution here. AI-based instant translation could be used to facilitate the initial conversation between the examiner and parents. In addition, such an LLM could help during language diagnostics by transcribing the child's responses in the language of origin so that they can be analysed automatically afterwards.

In our pilot project, we want to investigate how parents of Turkish-German-speaking children evaluate the use of AI in the initial interview and how reliably AI-based automatic speech recognition works in the evaluation of Turkish language skills.

If the initial results are promising, the method will be trialled in a larger study. Our long-term goal is to break down language barriers in language development diagnostics and enable a more valid assessment of the language development of multilingual children.

Project management
Dr Birgit Ehl

Project collaborator (University of Wuppertal)
Hatice Ekici Former research assistant

Duration: since 09/2024

Funding
IfB Innovation Fund