Evaluating Events in Narrative Theory

Evelyn Gius & Chris Biemann

The proposed project ‘Evaluating Events in Narrative Theory’ (EvENT) is an interdisciplinary approach to the formalization and analysis of events in narratives, based on a combination of Natural Language Processing (NLP) methods and concepts of Literary Studies. With a focus on the potential added value of using concepts of one discipline in the other one, the project includes the theoretical research of event concepts in both disciplines, the subsequent modeling and implementation for the automatic analysis of narrative texts, and the application of these systems for literary text analysis as well as in the context of information representation. For the NLP research community, the theoretical discussions on narrative research questions and the operationalization of respective concepts in cooperation with Literary Studies will lead to advances on modeling events, sequences of events and a notion of event centrality/event ranking within long documents, with the additional challenge from the application domain, i.e. German literary texts. For Narratology the modeling of events in cooperation with NLP will establish a testbed, which has not been available so far. Inductive development and application of categories for text analysis as performed in Literary Studies is complemented by a deductive approach, evaluating these categories on a variety of texts. This will lead to a twofold advancement of Narratology: with regard to theory the iterative modeling will generate theoretical findings about the formal narrative concept of events, with regard to the application of event concepts tools will be developed that can be applied for the computer-aided analysis of literary texts. The automation of the analysis of events and event patterns in literary texts will address three areas, which align with the central role of events in narratology: (1) the formal description of events as well as narratological models of narrative constitution based on them, (2) the analysis of event- fulness, which will be based on the event II concept and the results of the first item, and (3) the operationalization of both for the automatic analysis and classification of narratives as well as the analysis of further narratological phenomena with respect to the specificity of related event types or patterns.