Thomas Weitin, Katharina Herget, November 2016
Topic modeling is one of the more promising quantitative procedures for exploring semantic structures. The creators of the corresponding algorithms use large text sets to investigate hidden thematic connections which cannot be perceived by the eye alone.
We have, by contrast, tested a medium-sized corpus of novellas that can be explored using both individual readings and statistical procedures. We were motivated by a previously little considered observation: The scholars who have been able to implement statistical procedures for literary corpus analysis in a pertinent way were all extraordinarily familiar with their respective corpora.
Because we were also very familiar with our set, it was possible for us to order the topics being studied according to text-relevant themes. Using the keyword “falcon topics,” we describe another type of Topics, ones which seem to reflect the special character of individual texts.
[ Download (ENG) ]
[ Download (DE) ]