C11 - Prominence in Large Language Models
In this project, prominence is investigated with the help of large language models. Language models are to take on the role of test subjects in psycholinguistic experiments, which means that the project also pursues a methodological goal. Initially, the aim is to replicate the results of linguistic experiments conducted with humans as well as possible - whereby the project also has to work out how exactly this can be measured and what “as well as possible” means. To this end, we are working together with three projects from the CRC, namely B04, C04 und C07. The experimental simulations will then be gradually expanded so that larger contexts with more decisions and variants can be investigated. Finally, the project will also investigate whether language models can not only simulate a -- relatively undefined -- entirety of speakers, but can also be used to simulate individual test subjects.