Carolin Schuster

M.Sc.

Research Interests

• Interpretability of Large Language Models
• Semantics of Word Embeddings
• Utilization of Word Representations for Social Research, with a special focus on Topics, Attitudes & Bias
• Tokenization & Compositionality

Supervision of Theses / Guided Research

If you are looking for a thesis or guided research topic and are motivated to work on a project related to my research interests, don't hesitate to contact me via email! Please attach your CV, Transcript of Records, and a short motivational statement (< 400 words).

I am currently looking for students to explore the topics below. Applications will be reviewed after August 31, 2023. Please state your specific motivation for the topic of your choice.

• [Master Thesis] Relationship of Bias in Pretraining, Finetuning and Task Data - In this project you will use different methods to analyze bias in pretrained and finetuned LLMs, and investigate the relationships
• [Guided Research] Segmentation and Plausibility of Feature Attribution Methods for NLP - In this project you will analyze the impact of different segmentation choices on the plausibility of explanations by feature attribution methods
• [Guided Research] LLM-assisted Construction of a Corpus on Sustainable Product Reviews - This project entails two things: Creating the corpus and optimizing the process of its creation with LLMs
• [Guided Research / Master Thesis (scope will differ)] BERTitude: Exposing User Attitudes in Language Model Embeddings - In this project you will modify the BERTopic pipeline to reveal "topics" concerning specific entities/products