Investigating the Applicability of current Machine-Learning based Subjectivity Detection Algorithms on German Texts

In the field of subjectivity detection, algorithms automatically classify pieces of text into fact or opinion. Many different approaches have been successfully evaluated on English or Chinese texts. Nevertheless the assumption that these algorithms equally perform on all other languages cannot be verified yet. It is our intention to encourage more research in other languages, making a start with German. Therefore, this work introduces a German corpus for subjectivity detection on German news articles. We carry out this study in which we choose a number of state of the art subjectivity detection approaches and implement them. Finally we show and compare these algorithms’ performances and give advice on how to use and extend the introduced dataset.