Weiß, J., Marquardt, J., Meurer, J. E., & Eschenwecker, S. (2023). Das Codierschema des "wichtigsten politischen Problems in Deutschland" der German Longitudinal Election Study. (GESIS Papers, 2023/09). Köln: GESIS - Leibniz-Institut für Sozialwissenschaften. https://doi.org/10.21241/ssoar.89818
In surveys, questions are often asked openly, e.g. to record trends. At the same time, processing the resulting data involves a lot of effort. A typical example is the question about the currently most important problem, which is also regularly asked in the German Longitudinal Election Study.
In this context, the authors have now tested a new way of coding the open answers and published the resulting data product! Study number ZA7957 (https://search.gesis.org/research_data/ZA7957) now contains almost 400,000 open-ended responses, which were coded into the GLES coding scheme using the BERT large language model.
In the associated GESIS paper (https://www.ssoar.info/ssoar/handle/document/89818), the authors also describe the procedure leading to this data product. By using the Large Language Model BERT to code almost 400,000 open responses from the GLES, they have not only found an innovative approach for the efficient processing of large amounts of data, but have also created significant added value for survey research. This novel method not only enables time-saving and precise coding, but also opens up new perspectives for the analysis and interpretation of open-ended questions in surveys, allowing a deeper understanding of trends and developments to be gained.