For Students

Open Thesis Topics

Students who are eager to develop their skills by doing a research-oriented thesis in our group should mail their interests to webis@listserv.uni-weimar.de. Suitable topic candidates are shown in the following list. Your own suggestions for topics are also welcome, for which you can draw inspiration from our recent publications.

  • A Search Engine without a Server: Evaluating different Efficiency/Effectiveness Trade-Offs
  • Efficient Neural Retrieval with Splade-Doc on a Raspberry Pi
  • Extracting Causal Knowledge by Reproducing k-CNN
  • Human Values in Retrieval Augmented Generation Responses
  • Investigating Unconventional Automatic Differentiation Methods for LLM Fine-Tuning
  • Large-scale Rank Fusion Evaluation
  • The Critical Friend Paradigm in Watermarked Conversations with a Search-As-Learning Chatbot

Open Student Assistant Topics

Students who want to improve their skills and work with us can apply for a position as a student assistant at webis@listserv.uni-weimar.de. We are currently looking for assistants to work on the following topics:

  • We currently have no open topics. You can always let us know if you're interested.

Ongoing Theses

  • Jena
    • Evaluating Pre-Training Techniques for Single-Vector Encoder Models (supervised by Ferdinand Schlatt)
    • Testing the Limits of Multi-Vector Bi-Encoder Models (supervised by Ferdinand Schlatt)
    • Dense Boolean Retrieval for Systematic Reviews (supervised by Ferdinand Schlatt)
    • Improving Learned Lexical Retrieval Models by Removing Lexical Dependencies (supervised by Ferdinand Schlatt)
    • Reducing the Size of Dense Retrieval Indexes by Removing Unimportant Terms (supervised by Ferdinand Schlatt)
    • Key Point Generation for Different Datasets (supervised by Ines Zelch)
    • Construction of Fine-Grained Retrieval Pipelines With PyTerrier (supervised by Maik Fröbe and Jan Heinrich Merker)
  • Kassel
    • A* for Causal Inference (supervised by Tim Hagen)
    • Discrete Directions in the CLIP Embedding Space (supervised by Niklas Deckers)
  • Leipzig
    • Facets of complexity in scholarly political language (supervised by Magdalena Wolska)
    • Simplifying the language of political argumentation (supervised by Magdalena Wolska)
    • Lightweight Passage Re-ranking Using Embeddings from Pre-trained Language Models (supervised by Ferdinand Schlatt and Harry Scells)
    • Logical Features of Neural Networks (supervised by Maximilian Heinrich)
    • Active Learning for Text Classification (supervised by Christian Kahmann and Christopher Schröder)
    • Incorporating Knowledge Graph Embeddings in Large Language Models (supervised by Ferdinand Schlatt)
    • Normdaten-Disambiguierung und Reconciliation auf Korpusdaten (supervised by Erik Körner and Felix Helfer)
    • Statistical Bootstrap Tests with Redundant Data (supervised by Maik Fröbe)
    • Mining Trigger Warnings from the Web and Social Media (supervised by Matti Wiegmann)
    • Web Search Archeology in the Archive Query Log (supervised by Jan Heinrich Merker, Simon Ruth, and Matti Wiegmann)
    • AQLQA: Mining Direct Answers from Dozens of Search Engines over 25 Years (supervised by Jan Heinrich Merker, Simon Ruth and Maik Fröbe)
  • Weimar
    • User Simulation in Persuasive Dialogs (supervised by Marcel Gohsen and Nailia Mirzakhmedova)
    • Search Session Detection in the Archive Query Log (supervised by Marcel Gohsen and Jan Heinrich Merker)
    • Employing Personal Knowledge Graphs in User Simulation in Conversational Search (supervised by Nailia Mirzakhmedova)
    • Building an Image Generator for Arguments (supervised by Maximilian Heinrich)
    • Information Extraction from Scientific PDFs (supervised by Tim Gollub)
    • Is this sound? Mining and Evaluation for Argumentative Fallacies (supervised by Maximilian Heinrich)
    • Adding Contextual Awareness to LLM-based Story Generation (supervised by Tim Gollub)
    • Prompt Framing Effects on Large Language Model Subjectivity Judgements (supervised by Nailia Mirzakhmedova)
    • How Humans Detect Cloned Voices (supervised by Johannes Kiesel and Marcel Gohsen)
    • Crafting Multimodal Learning Experiences based on Course Materials (supervised by Marcel Gohsen)
    • Efficient and Effective Neural Translation Language Model for Search (supervised by Harry Scells)
    • Rating the Degree of Search Engine Optimization of Websites (supervised by Janek Bevendorff)
    • Re-ranking with Health-related Retrieval Axioms (supervised by Jan Heinrich Merker and Maximilian Heinrich)
    • What is the argument of the day? Systematic Mining of Arguments (supervised by Maximilian Heinrich)

Resources for Students

Vacancies

Dear prospective PhD student, unsolicited applications to the Webis group (webis.de) are welcome. However, we cannot promise that open positions are available at the time of your application.

The Webis Group is a tightly cooperating research network, formed by computer science chairs at the universities of Groningen, Hannover, Jena, Kassel, Leipzig, and Weimar. Our mission is to tackle challenges of the information society by conducting basic and applied research with the goal of prototyping and evaluating future information systems. We are an experienced research group where team spirit and active collaboration has top priority. We are looking for open-minded graduates and PhDs who want to develop both as a researcher and as a person. The working language of our group is English; fluency in German is not required.

Interested students should have finished either a master or a PhD in computer science, mathematics, or a related field with excellent or very good grades. A solid background in mathematics and statistics is expected—as well as very good programming skills.

Benno Stein
Bauhaus-Universität Weimar
On behalf of the Webis group

Email: webis@listserv.uni-weimar.de
Web: webis.de