The Third Workshop on Computational Models of Narrative

I discuss a collaborative, computer aided approach towards building and exploiting a shared resource that can aid further research into the history and development of narrative, as well as into its phenomenology and logic. The particular example illustrating this approach is a project called CLÉA, short for Collaborative Literature Éxploration and Annotation. CLÉA, a Google Digital Humanities Award funded project1, is a browser based annotation and text analysis environment which comprises three functional modules: 1. a working environment for highly unconstrained, nondeterministic collaborative markup of narratives; 2. a repository which manages and distributes object data (texts and corpora, imported among other from Google Books), and meta-data (stand-off markup files) as well as tag sets (consisting of TEI-XML compliant narratological tags) aggregated and generated by users; 3. a heuristic module, based on a machine learning component, that aims to identify and draw the human user’s attention to hitherto unnoticed patterns and regularities in the phenomenology of human-made narratives. In the field of narrative studies, whether philological or computational, CLÉA’s crowd sourcing approach is novel in at least two regards: One, it aims at basing our analyses, theories and models of the phenomenon of narrative representation on collaborative corpus studies, rather than on the detailed, in-depth investigation of a small number of exemplary narratives that is undertaken by an individual researcher. Two, it augments human research activity by a computational heuristic that works bottom-up, through the statistical analysis of human generated metadata, rather than top-down, i.e. by mapping a pre-defined abstract model or taxonomy of narrative onto the original object-data itself. In doing so, this heuristic acknowledges the crucial fact that narratives are always human constructs, CLÉA is the current, second phase development of the initial stand alone desktop application CATMA (Computer Assisted Textual Markup and Analysis) conceptualized and developed at Hamburg University from 2009 onward. In heureCLÉA, its third development phase which we plan to commence in 2013, we will extend system support for narratological markup and implement a robust machine learning based heuristic module. For more technical information on CATMA and CLÉA as well as downloads, see http://www.catma.de. and not real-world entities that can be found ”out there” and on their own. However, CLÉA amounts to more than merely facilitating a new praxis of analysing and modeling the symbolic processes and products that human cultures designate as ‘narrative’. In a methodological perspective and with regard to the current transformation of the humanities in general, CLÉA also demonstrates the impact and potential relevance of the new scientific paradigm of the Digital Humanities. In a disciplinary perspective on the other hand, and in particular in that of narratology and of computational science, CLA may be considered as an example for an emerging inter-discipline, tentatively labeled by Inderjeet Mani and others as Computational Narratology.2 Accordingly, I will pay equal attention to the demonstration of CLÉA as a concrete example, and to the reflexion of broader methodological and programmatic consequences. Adopting a computer aided, crowd sourcing based approach in the study and modelling of narrative, I believe, will do more than afford us new insights into the logic of narratives as objects of our research. It will also help us to understand better on which premises our reasoning about narrative is based, and which contingencies and constraints — disciplinary as well as cultural — we might have to take into account in our future work. See Mani’s forthcoming article ”Computational Narratology” in Hühn, Peter et al. (eds) The living handbook of narratology. Hamburg, Hamburg University Press. http://hup.sub. uni-hamburg.de/lhn