Machine Learning (ML) delivers success across domains like autonomous driving, cancer detection, ornatural language processing. However all established methods need copious amounts of labeled training data to learn. Labeling is a „costly, error-prone, and labor-intensive“ process gernally performed manually. Hereby interactive intelligent labeling (IIL) systems are used to optimize this task. They offer an increasingly big design space, and research in this specific field has been increasing exponentially in the last years.

Scholars and practicioners alike lack a structured overview of the state-of-the-art, along with prescriptive knowledge about how to design such systems. This thesis is thereby framed as part of an ongoing research project and offers the rare opportunity to contribute in a significant way.

1. Coding of a previously conducted structured search of the related literature, over a total of 197 papers. Hereby, the overarching themes of the morphological box have been already derived top-down, whereas individual categories have to be developed bottom-up.
2. Transformation of the resulting morphological box into a web-based recommendation system / configurator. Hereby this tool should allow for e.g. selecting a certain task type, and then output a recommended system configuration.

The applying student is expected to be familiar with the concept of qualitative and quantitative literature reviews. Experience with manual coding in such contexts is beneficial. We further expect the student to be familiar with web development. 

If you are interested in this topic and want to apply for this thesis, please contact Merlin Knäble with a short motivation statement, your CV, and a current transcript of records. Feel free to reach out beforehand if you have any questions.