Workshop - Large Language Models (LLM)

“Bag-of-words” approaches or frequency analysis are (outdated) text analysis methods that hardly do justice to the complexity of natural language. For this reason, deep learning or neural networks are innovative, important tools for researching large amounts of text that can help overcome these limitations. Moritz Edlinger, a PhD student at the Digital Psychology Lab who is working on the analysis of user-generated reviews in his project, therefore attended the workshop “From Embeddings to LLMs: Advanced Text Analysis with Python” as part of the GESIS Fall Seminar workshop series. During the five-day workshop in Mannheim, Large Language Models such as BERT, BERT or OLLAMA were dealt with in detail and their possible applications in everyday research practice were discussed. Advantages and disadvantages of the respective models, as well as techniques for fine-tuning the models, were also a central topic of the workshop. Moritz describes his time in Mannheim as “instructive” and was glad to have attended the workshop. Especially for first year PhD students, workshops like this are a great opportunity to get up to speed with the latest technology and at the same time meet other researchers from the same field.