Interview with Alexander Frummet

I work at the Chair for Information Science (Institute for Information & Media, Language & Culture).

So far I have given basic lectures on the IW / Media Informatics (saving and processing data efficiently, IR advanced seminar, case study seminar), but also taught courses for the DH course (e.g. basics in Python, web technologies). A special focus cannot be identified here, but in the DH Master mainly lectures on development / programming topics.

In the course “efficient storage and processing of data” one learns among other things how to properly prepare data, e.g. avoid redundancies. In addition, yu learn how to access this data using SQL. Python (DH master’s course) is also an important language used for data processing. Finally, in web technologies, one learns how to use data e.g. Django / Javascript / HTML virtualized and suitably prepared for the user. Most of the reference to AI /ML can probably be made with Python. This language is important in connection with data science and machine learning, as many machine learning libraries (Keras, Tensorflow, Pytorch…) work with Python.

Since my courses provide the basics or “tools” with which you can later do machine learning/data science, I do not have to fall back on any particular prior knowledge

In the python course, where I led the exercise together with Manuel Ullmann, you learn, for example a lot about data structures and how to get data in the right format so that it can be processed by a classifier.

In my opinion, AI / machine learning and data science will be even more in-demand skills on the job market in the future than they are now. It is often said that “Data is the oil of the 21st century”. Being able to handle the abundance of data, structure it and, above all, be able to derive the right decisions from it, is an important skill.

In my research I deal with conversational search and the wuestion of how to make systems understand user utterances better.

I used different ML classifiers for the automatic recognition of information needs in user utterances.

I studied general and comparative linguistics and information science in my bachelor’s degree and media informatics in my master’s degree. I was introduced to the contet through information science and media informatics, bu I also did a lot of self-study in this direction.

At the beginning it was not so easy for me to understand how certain algorithms and neural networks work mathematically. Over time, however, that became clearer and clearer. From an application perspective, there are easy-to-use libraries like Keras and Scikit. This makes handling particularly easy.

When machine learning works, it works like magic. The math behind how most algorithms work is relatively easy to understand, which makes the whole thing even more remarkable.

There are currently no interdisciplinary collaborations.

Ai methods should be seen as a means to an end and a tool to answer different new questions. ML methods can be used in many subjects, e.g. also in the humanities. This certainly gives rise to many interesting research questions. The DH program shows that.

“KnowledgeCreatesData” is an important cross-faculty project. It is important that subjects work together across faculty boundaries. So far, my contribution has been to help lay the foundations for data analysis and processing (see Python)