From Networks of Texts to Networks of Topics? On the Classification of (Texts in) Compilations with a View towards Manuscript Transmission

As medieval manuscripts often consist of more than one text, the application of network analysis can show textual connections between codices and therefore shed light on the circulation of texts, of manuscripts, and thus of knowledge. However, a text-based analysis often faces difficulties resulting from insufficient manuscript descriptions and a lack of normalization of work titles. A broader view, which would compare not particular texts, but rather genres, areas of interest or fields of knowledge, may help to circumvent these problems; however, this broader approach must deal with problems regarding classification. Instead of finding connections between subjectively classified texts, one can make use of topic modeling as a means to computationally classify, and thus characterize, multiple-text manuscripts. On the basis of automatically detected topics, topic-based networks can be generated. The current potential of such an analysis was tested using a sample of codices that contain the late medieval chronicle of Jakob Twinger von Königshofen. Advancements in text recognition and normalization of non-standardized spelling could further enhance this method to investigate the connections between the codices of a specific corpus and develop a better understanding of the copying and transmission of premodern manuscripts.