Time after time: Representing time in literary texts

Abstract

The representation of temporal information in text represents a significant computational challenge. The problem is particularly acute in the case of literary texts, which tend to be massively underspecified, often relying on a network of semantic relations to establish times and timings. This paper shows how a model based on threaded directed acyclic graphs makes it possible to capture a range of subtle temporal information in this type of text and argues for an onomasiological approach which places meaning before form.