How to apply

If you are interested in completing your bachelor’s or master’s thesis with our group, feel free to get in touch with your research proposal via amllab[at]bit.uni-bonn.de or contact the relevant researcher listed on our team webpage. Each researcher’s page includes their areas of interest. You can include your CV, transcript of records, and links to any relevant software projects you have worked on, if available.

Research Proposal

Your research proposal is a central component of your application. It outlines the research question you aim to address and gives you the opportunity to demonstrate your expertise in the subject area, as well as the methodological approach you intend to take.

We review proposals to assess how well your research interests align with those of our lab and its supervisors. If you are interested in working with a particular supervisor at our lab, please indicate this in your proposal.

A strong proposal should:

• Clearly state the research question you want to investigate.
• Provide relevant background and context for your chosen area.
• Explain the novelty and significance of your proposed research.
• Describe the methods you plan to employ.

Feel free to use this resource from the York St. John University for inspiration: https://www.yorksj.ac.uk/study/postgraduate/research-degrees/apply/examples-of-research-proposals/

Procedure

Our lab follows several specific procedures when it comes to bachelor’s and master’s theses, namely:

• The master thesis usually takes about 6 months (it is planned for about 20-24 weeks)
• The working language of our group is English. That means, the meetings with the mentor and supervisor are usually in English, and the presentations to the team as well. Ideally, the work is also written in English, so that the students can also get regular feedback from the mentor while writing.

Furthermore, there is a fixed structure for meetings with the lab’s members:

• During the first meeting, the student pitches their idea and receives feedback on its feasibility.
• 2–3 months later: Progress meeting, including additional feedback and a brief progress report
• Another 2–3 months later: Final progress report.
• After the thesis is submitted: A defence date will be scheduled as soon as possible.

Additional meetings can be scheduled at the discretion of the mentor or supervisor.

Topics

It is usually most effective to propose a topic that genuinely interests you, as this can help maintain motivation throughout the project. This approach also more closely resembles the way research is conducted in practice: while researchers often have guidance or general themes, they are generally expected to develop and refine their own research questions.

If you find it challenging to come up with a topic of your own, don’t worry. Below is a list of suggested topics that may serve as inspiration or a starting point. You are welcome to directly use these ideas, combine or adept them, or use them to help formulate your own research question in your thesis at our lab.

List of Topics

This list of topics is organized by the mentor that is offering to supervise the thesis. If you think one of these topics is of interest to you, contact she/him directly (see the Team page for contact details).

Farizeh Aldabbas

• Multilingual and Cross-Lingual Fact-Checked Claim Retrieval
  • Develop a multilingual retrieval system that identifies relevant fact-checked claims for a given query claim. The project focuses on semantic similarity using multilingual embeddings and ranking methods, and analyzes how retrieval performance varies across languages.
• Multimodal Financial Claim Verification
  • Build a system that verifies financial claims using both textual and visual evidence. The project explores how text extracted from financial images (e.g., charts or screenshots) using OCR can be combined with claim text to improve automated financial fact-checking.
• Psycholinguistic Conspiracy Marker Detection
  • Investigate the detection of conspiracy-related narratives in online text by identifying psycholinguistic markers associated with conspiratorial discourse. The project focuses on developing classification models to detect such patterns and analyzing the linguistic signals of conspiracy rhetoric.

Tobias Deußer

• Constrained Entity Decoding with Diffusion Models
  • Mandatory reading before contacting: Large Language Diffusion Models by Nie et al., Informed Named Entity Recognition Decoding for Generative Language Models by Deußer et al., Diffusion-LM Improves Controllable Text Generation by Li et al.
• Contradiction Detection in Long Documents
  • Mandatory reading before contacting: Uncovering Inconsistencies and Contradictions in Financial Reports using Large Language Models by Deußer et al, ContraDoc: Understanding Self-Contradictions in Documents with Large Language Models by Li et al.

Hossam Elsafty

• Diacritization (Tashkeel) Generation and Downstream Arabic NLP Effects
• MSA to Dialectal Arabic NLP tasks
• Arabic Document Understanding and OCR

Corinna Schmalohr

• Benchmarking Large Language Models for Clinical Documentation: Accuracy, Bias, and Safety in AI Medical Scribes
• Design and Evaluation of AI-Assisted Clinical Decision Support Systems for Rare Disease Diagnosis
• Robust Information Extraction from Noisy Clinical Documents: Combining OCR and Language Models
• Hallucination Detection in Medical Large Language Models

Svetlana Schmidt

• Design and evaluation of a food and nutrition ontology for integrating food data sources.
• Ontology-aware semantic retrieval for ingredient normalization in food product data.