Stories have structure. The fundamental elements of narratives are events and causal relationships that reference concepts, entities and people. By structuring these elements using event and causal ontologies it is possible to capture narrative information in computers, enabling navigation, reasoning and other operations and driving innovative narrative-based applications. Non-fiction and news narratives are not structurally different from fictional narratives, however non-fiction narratives can be connected into a narrative network, enabling much more extensive navigation, inference and reference operations. This representation of structured narratives within a narrative network is at the core of the StructuredStories technology.

The research basis for an analytical approach to narrative has developed over the last 60 years in multiple fields. Until recently it has been dominated by theoretical work in university humanities departments – a field known as Narratology. Other fields have recently begun making substantial contributions to our understanding of narratives, including cognitive neuroscience, evolutionary psychology, causal modeling, computational linguistics, computer science and artificial intelligence. Recent applied work on computational narrative has also been done in journalism labs, in game development labs, in entertainment production companies and at several start-ups. Some significant recent developments in this emerging field include the successful commercial use of case-based reasoning to automatically generate narrative from structured financial and sports data, and the early use of ‘structured publishing’ by several global news organizations. Analytical and computational narrative is multidisciplinary to an extreme degree – a fact that has slowed the emergence of innovation and useful applications of narrative in digital media.

The design of the StructuredStories narrative platform is based on integrating these diverse perspectives on operating with narratives in computers, and on applying this integrated view of the research to emerging demands in digital media – especially in media and information personalization. New tools such as graph databases, semantic web standards, various open source frameworks, knowledge services and publishing technologies combine to make modeling and testing these techniques with live applications much easier. The StructuredStories narrative platform seeks to use this integrated research, these digital media use cases and these new technical tools to deliver core functionality in four areas:

  • Narrative Capture

The most significant challenge in building useful narrative systems is in capturing structured events and their related entities and causal links from the environment in sufficient detail and at sufficient scale to be useful. There are successful examples of both automated and manual approaches to this challenge with advantages and disadvantages to each. The StructuredStories platform  combines both automated and manual methods into a single capture process that enables managed trade-offs between detail and scale and that facilitates both professional and crowd-sourced narrative capture.

  • Narrative Knowledge Representation

The heart of a computational narrative system is the set of data structures in which narrative information is stored – specifically event, entity and causal linkage information. The StructuredStories platform uses a graph database for this representation, expressing custom event and causal ontologies and accessed by capture, inference and query tools using a REST API. An external knowledge graph is referenced, and existing semantic web and ISO standards such as TimeML are used extensively.

  • Narrative Inference

Inference and reasoning on narratives takes advantage of structural, temporal and causal logic to derive new knowledge and insights from narratives and from networked collections of narratives. The potential power of narrative inference is considerable, especially in a richly-populated networked narrative environment, however narrative inference is still a very nascent concept. The StructuredStories narrative inference engine is initially focused merely on validation, on simple capture pre-processing and on simple query-related operations in support of applications, but it also provides a platform for the future development and testing of more advanced narrative inference techniques.

  • Narrative Applications

The purpose of the StructuredStories platform is to enable rapid experimentation in digital media applications using narratives.  Applications can use the StructuredStories API to originate narratives and to capture narrative elements within them, as well as to search for and retrieve existing narrative information. Applications can also use several modules for basic navigation and display of narratives – for example navigation through connected events at successive levels of narrative detail via the fractal narrative module. The StructuredStories API and application modules will be used for two news-related demonstration applications during 2014, and should be available for access by external applications on a limited basis by late 2014.

As an applied research and development company, StructuredStories is primarily engaged in designing and building this core functionality and applying it to digital media use cases to the ‘proof of concept’ level. This effort includes the development of associated intellectual property and includes development of a market-oriented understanding of the actual response of digital media consumers to narrative applications. Commercialization of the technology is not a primary objective at this stage.