Information Technology Reference
In-Depth Information
system-building experience and on the four master tropes identified by semiotic
research, namely: metaphor, metonymy synecdoche, and irony (which are the basic
rhetorical structures by which we make sense of experience [2, 5]). By offering
mechanisms derived from event relations connected to these tropes, we intend to
augment the expressiveness of the narrative model.
This paper describes a conceptual model used to represent plots in a given genre
and presents extensions to better support event relations and enhance the space of
possibly interesting stories in the context of a plot generation system. We outline how
an intended genre (to whose conventions the plots must conform) can be modelled
and describe LogTell-R [10], an extended version of LOGTELL [4]. This new system
uses
plan recognition
and
plan generation
in the creation of stories and a
plot
algebra
to help define the story space. We then show how the new system works on
the basis of the identified event relations, also indicating what is done to narrow the
gap between the conceptual model and the implementation while helping to construct
the possible plots. Finally, we offer some remarks on the results obtained.
2
Story Generation Systems
Story generation systems (SGSs) employ different strategies to configure their story
models, which can be broadly classified as character-based, plot-based, and user-
experience models. In character-based models, the storyline results from the
interaction among virtual agents. Their main advantage is the ability for anytime user
intervention, but, while powerful in terms of interaction, such interference level may
lead the plot to uncharacteristic situations. Additionally, there is no guarantee that
narratives emerging from the interaction between agents will create interesting drama.
By contrast, in plot-based models, characters should follow more rigid rules,
specifying the intended plot structure. User intervention is thus more limited, but it is
usually easier to guarantee coherence and a measure of dramatic power. Lastly, SGSs
with user-experience models, focus on the user by addressing individual preferences
via user profiling and measuring the interestingness of story pieces.
We feel that an approach that fits in between plot-based and character-based is the
best option, given that plots and characters are interlocking elements that cannot exist
without each other [7]. Specifically, we try to conciliate both. User-experience
concerns can be dealt with at later stages; as long as we can guarantee that enough
variety of stories can be generated, possibly while considering changes in
dramatization / presentation of story events. Even though it is argued that the key
aspect of an interactive narrative is the story representation used to encode the
author's vision of the experiences (story space [12]), little attention is given to what
qualities such space should have. Most approaches to plot-influenced storytelling
models consider only stories formed by sequences of events using their inherent
temporal characteristics or simple causality relations, which are clearly not enough to
guarantee interesting and varied stories. Other approaches use dilemma-inducing or
tension-raising events, paying no attention to the relations between events in the plot
model. Few efforts dedicate enough attention to relations between events. Three
