Information Technology Reference
In-Depth Information
Support for complex operations in plan recognition and the use of the hierarchy of
typical plans can be seen as a complement to generation. The benefits of the plan
library structure itself are twofold: it helps better visualize story structure, showing
how to adequately chain events into a narrative scheme (like the three-parts classic
structure [10]); and it helps in more easily understanding and analyzing the space of
stories that can be generated. In LogTell-R, plot generation still happens in a step-
wise fashion. A new way of choosing alternatives is enabled by selecting events and
asking the plan-recognizer to retrieve generalizations containing them. Exploiting the
meronymic axis at different levels of detail, plausible alternatives may be located by
essentially walking up and down over the hierarchy. Instead of asking the system for
suggestions, the user may also simply find it helpful to visually inspect the library of
typical plans associated with the genre (Figure 2), and use it as inspiration to guide
the composition of a plot.
As our approach is centered on the concept of hierarchies of typical plans and
given that conceptual models tend to remain too far removed from system
implementations, we narrow this gap with the help of an intermediate logical design
stage. The approach utilized is a plot manipulation algebra (PMA) [9] for the creation
of the hierarchy of typical plans. An added benefit of using PMA (supported by the
planner to check pre- and post- conditions of operations) in the knowledge base editor
is that one can more easily experiment with different ways to chain the events, and
use this knowledge to revise the specification so as to improve the system's story
space. Starting with the set of operations relevant for the genre (e.g. abduct, murder,
liberate), algebra operands are applied to define the necessary relations between
events (e.g. generalizations or placement of constraints on undesired event
combinations). At each step, the user can visualize how the typical plan hierarchy is
“grown”.
6
Concluding Remarks
Few approaches to IS address the problem of what is a proper story space or the
characteristics of events pertaining to a plot, preferring to focus on more controlled
environments and stories. We feel that more generative approaches are necessary for
the field to become truly successful. It is important to stress, however, that we do not
propose to achieve a fully immersive experience. We endeavour, instead, to explore
the possibilities of generating a large variety of coherent and interesting stories
through the use of plan-generation and recognition over the conceptual model of a
genre.
Our approach targets sound methods to organize and combine events, in ways that
may confer enough dramatic power to the generated narratives. By extending our
conceptual model, we augment its expressiveness and provide tools to prospective
authors intent on creating and telling stories. Although the process of plot
composition and adaptation could surely be enriched beyond what is presented here,
our system provides a sound basis to treat genres that exhibit a good degree of
regularity.
