Theses
We offer currently theses (diploma, internship, bachelor, and master) in the following fields:
- Data Knowledge Engineering, Data Mining, Machine Learning
- Information Retrieval
- Computer Linguistics
- Music Retrieval
- Visualization and User Interface Design
The thesis can be written either directly at our group or at one of our cooperation partners. Specific topics can be get from the members of our research group. A small selection of possible topics is presented in the following.
Hints on how to write your thesis can be found here.
A Small Selection of Available Topics
- Analysis and Structuring of fMRI Data
Contact: Prof. Andreas Nürnberger, Prof. Klaus-Dietz Tönnies
- Design of a User Centered Interfaces for a Personalized Search Engine
Contact: Marcus Nitsche
- Comparison and User-Oriented Study of Different Visualizations of Hierarchically Structured Search Result Sets
Contact: Marcus Nitsche
- Novel User Interfaces for Searching Through Eyetracking, Wii-Controller oder iPod touch
Contact: Marcus Nitsche
- Computation of a (web) document complexity/readability score
Different documents contain information with different complexity levels. Documents which are easy to understand for adults could be inappropriate for children and vice versa. Within this work you should research how the (web) document complexity level can be derived and develop a model that computes complexity score (or a set of scores) for a given document. The validity of the scoring approach should be evaluated, ideally based on a prototypical implementation written in Java.
Contact: Tatiana Gossen - Evaluation of overlapping clustering techniques
There already exist quality indices for graph clustering as coverage, performance, intra- and inter-cluster conductance, modularity etc. In this work you should modify the existing indices for the case of overlapping clustering and prove their correctness and properties. Then you should implement one of the existing overlapping clustering algorithms and do experiments to measure the quality of your clusterings with the new indices based on benchmark datasets.
Contact: Tatiana Gossen
