'No results or wrong'? - Methodological Challenges in Computational Literary Studies

Computational Literary Studies has developed rapidly in recent years. The existence of larger text collections in some languages as well as the easily accessible tools to generate interesting research results with these collections has contributed to this. Although some parts of computational literary studies can draw on longer traditions, in many respects they are a result of recent years in which many approaches, methods and tools have been tested. The lecture will present some of them, but above all it will examine the question of whether this field has specific methodological problems, as some critics think. This criticism is often based on the assumption that literature is a particularly complex subject that cannot be dealt with by quantitative methods. But even if one admits that literature is probably not more complex than the human psyche or human societies - which are each examined by sciences that work intensively with quantitative methods - the question arises as to how the methods of CLS prove themselves in methodological testing.  What role can explorative methods play in knowledge acquisition, for example, and where are their limits? Or how can approaches to machine learning, which are known to follow optimization logic, be robustly and fruitfully combined with inferential statistical approaches? The lecture will discuss these problems without being able to offer complete solutions right away, but will perhaps include some warning signs.