Kopräsenz-, Koreferenz- und Wissens-Netzwerke. Kantenkriterien in dramatischen Figurennetzwerken am Beispiel von Kleists Die Familie Schroffenstein (1803)

Abstract In my contribution, I argue that a fruitful integration of computational network analytic methods into literary studies depends on how the ›interaction‹ of two characters in an abstract character network is formalized. I support this hypothesis by using the examples of co-presence, co-reference, and knowledge networks, which I analyze in Heinrich Kleist’s tragedy Die Familie Schroffenstein (1803). I assume that the co-presence of characters can provide the basis for more specific formalizations of interaction. But due to its basal specification of interaction, co-presence networks can only be integrated into rather limited questions of literary studies. I illustrate this circumstance by examining Franco Moretti’s approach in his essay Network Theory, Plot Analysis (2011): How does he connect his network analyses to concepts of literary studies? How does he reflect his methods? Which observational stance does he adopt? How appropriate is his approach to the object of study? And how does he tie his results back to literary theory? Moretti’s explorations show that his network analyses seem to be incompatible with established conceptions of characters – at least partially. Therefore, he demands a new conceptualization of dramatic characters in literary studies. To me, however, it seems to be more productive to put into perspective or enhance – under the auspices of network analysis – existing quantitative aspects of established character presentation (configuration, constellation). I therefore propose two additional formalizations of character interaction to create dramatic networks: co-references and knowledge transfers. Regular co-presence networks, which have widely been tested and discussed among (computational) literary scholars, will serve as a ground of comparison. I illustrate the merits and limitations of co-presence networks on both a single text analysis of Die Familie Schroffenstein and a larger corpus analysis of 587 German-language plays. Cursorily, I present the operationalization of co-references and knowledge transfers. The linguistic concept of co-reference means that two or more linguistic expressions refer to the same entities. A knowledge transfer, in my understanding, is a transmission of new information from at least one literary character to at least one other character. Manual annotations of co-reference chains and knowledge transfers serve as basis for the subsequent network creation. I compare these different manifestations of character interaction in terms of the resulting network visualizations as well as of various mathematical network metrics. The goal is to elicit how useful these two criteria are regarding drama analysis, e. g., the analysis of character properties, and to what extent they can complement, differentiate, or even replace the established co-presence networks. Co-presence, co-reference and knowledge networks reveal different aspects of the characters under consideration, picture different dramatic structures, and place different groups of characters at the center of the networks. Therefore, the three abstract textual representations seem to complement each other. In my article, I show that the extent to which the various criteria can be integrated into research questions of literary studies ought to be discussed on (at least) two levels. Firstly, it is necessary to ask how interesting, relevant, and informative a specific criterion is for one’s own research and whether it can relate to the terminology of literary studies. Secondly, it is important to consider how precise the respective criteria can be annotated manually and, subsequently, if and how reliable they can be annotated automatically.