Metaphor mining in historical german novels: An unsupervised learning approach

This paper describes a work-in-progress to identify and categorize metaphorical language use in a large corpus of historical German novels. An unsupervised learning method is utilized to detect metaphorical expressions and underlying conceptual metaphors. Furthermore, an extension is proposed that allows for the analysis of diachronic developments of modeled metaphor types. A corpus ranging from the 16th to the 20th century serves to illustrate the challenges of this approach as well as its potential, not only as a tool for the analysis of stylistic variation, but also as a glimpse into the conceptual world views embedded in the texts under examination.