Chris Stahlhut, M.Sc.



Ubiquitous Knowledge Processing (UKP) Lab


Argument Mining, Interactive Machine Learning, Large Scale Dataprocessing


Interactive Machine Learning of Arguments and Argumentative Structures

Let’s say we want to see how the nuclear meltdown in Fukushima in 2011 and Chernobyl 1986 affected the political argumentation around nuclear energy and the following decisions. We might start by reading all political debates around the events, but that will only give us the immediate vicinity of the event and not the long term opinions and arguments and it is time and labour intensive. Extracting the arguments automatically seems to be a great idea, but there are no pretrained models for argument extraction on political texts available. This is where my PhD project starts. My goal is to take a machine learning model which has been trained to extract arguments from, say essays to let an exert user interactively improve this model so that it fits the experts purpose. Of course, I cannot assume the expert to have any knowledge about machine learning, hence I need to allow the human to teach in a way that humans find natural and machines can understand.


Studium & Ausbildung



Vorträge & Konferenzen