Unitizing Plot to Advance Analysis of Narrative Structure (PLANS)

Chris Biemann & Evelyn Gius

The proposed project “Unitizing Plot to Advance Analysis of Narrative Structure” (PLANS) is an interdisciplinary effort towards exploring and modeling concepts of plot in narrative theory by combining a theoretical perspective with an implementation-focused approach. As the starting point to these considerations, it is important to note that narrative theory lacks a comprehensive theoretical model of plot, just as comprehensive computational modeling of plot is infeasible with the current state of natural language processing techniques. Therefore, we seek to advance the analysis of narrative structure by working on an intermediate level: the modeling of plot units. With plot units we tackle the intermediate level between events and plot, or overall narrative structure. Thus plot units enable us to make substantial progress in the analysis of narratives, in narrative theory as well as in natural language processing.In PLANS, we will build on the event extraction in the predecessor EvENT project and adapt it in order to advance plot analysis. For this, we need to take into consideration that plot is determined both by its textual representation and properties of the fictional world, such as characters and events.Therefore, with the modeling of plot units, we strive to also conceptually connect the textual level of discours to plot, action, and generally the level of histoire; in simpler terms: we connect the how of narration with the what of narration.With this approach we strive to overcome current limitations in natural language processing and narrative theory. In natural language processing a variety of approaches to the extraction and understanding of plots have been taken, existing approaches, however, are not generally applicable to wide ranges of texts. An approach guided by narrative theory may enable us to create computational models of plot that are more broadly applicable, with the unitization-based approach enabling a more direct connection from the text’s surface from to its plot structure. Moreover, the segmentation based approach to plot units allows us to progress from the analysis of smaller, defined portions of texts to the entire narrative. This simplifies computational processing, enabling the use of otherwise impractical processing techniques (e.g., for coreference resolution or modeling of semantic world knowledge).With regard to narrative theory, the focus on the intermediate level between events and plot allows for a more general operationalization of narrative constitution, i.e., the principles that explain how events are combined and transformed into narrative text. This is still a desideratum in narrative theory. Moreover, we expect to gain insights into the as yet under-theorized conceptual connections between theoretical models of plot and narrative constitution.