Information Technology Reference
In-Depth Information
4.1
Genre as Story Space
Finding an answer to the question of what constitutes a given literary genre should
allow one to determine whether or not a story can be classified as belonging to it.
However proving that a conceptual model
M
for specifying a genre
G
, formulated as
discussed in this paper, can fully capture how the genre is understood by a literary
expert can be an over-ambitious effort. Yet, a useful approximation is attainable,
taking the form of a definition by extension. If a system can generate stories
according to the specified conceptual model
M
, the space of plots that can be
generated might be regarded as constituting a genre
G
M
defined by the model, where
plot has the usual meaning of sequence of events. It would be left to the designer of
M
the task of refining the model specification in order to achieve increasingly closer
matches between the intuitive view of
G
and
G
M
. This is basically the role of the
presented conceptual model. As argued, it determines the genre of the stories to be
generated in correspondence to the story space of a system that uses that model. Such
definition of genre as the set of stories that can be generated from a model is
consistent with the way game designers discuss/describe genres by example [19].
While the story space of our example scenario refers to a simple medieval fantasy
genre, our conceptual modelling method should be able to cope with an ample variety
of genres of higher complexity. The structure of each story space thus specified is
determined by the observed event relations. From an informal viewpoint (cf. Figure
1), events can be seen as nodes and syntagmatic, paradigmatic relations as connecting
edges drawn over a plane, while meronymic relations appear as projections of events
into event sequences over another plane. Antithetic relations between nodes can be
seen as specifying constraints crosscutting the story space. Recall that paths formed
by syntagmatic relations correspond to plots, whereas paradigmatic relations indicate
alternatives - whose choice may entail, as signaled by antithetic relations, the
exclusion of certain other alternatives branching from subsequent nodes.
Supporting the identified relations also brings extra benefits. Plot libraries can be
organized as combined is-a and part-of hierarchies, which is a convenient way to deal
with plots by taking advantage of the similarity or analogy among situations. While the
part-of links result directly from meronymic relations, denoting in consequence
composition / decomposition, the is-a links express the notion of generalization /
specialization. If two or more events stand in a paradigmatic relation, their main effects
should coincide, although they may differ with respect to less important side-effects. For
instance,
abduction
and
elopement
can be seen as alternative forms of
villainy
. This
justifies the introduction of a new event, which might be named
villainy
(or some more
specific term), and its placement in a hierarchy above the other two one-event plots, i.e.
it generalizes both. These relations contribute to the hierarchical structure of the library
and, once it is constructed, its component patterns can then be reused during the plot
composition process to help create new plots. Because our approach employs plan-
generation/plan-recognition while composing a plot, we regard “typical plots” in our
context as synonymous to typical plans. These relations also bring a particularly
convenient way to deal with entire plots, helping visualize their dramatic structure.
Finally, let us recall that we have addressed the fabula level only, where one simply
indicates which events should be included in the plots. One especially complex problem
to be faced at the next level - story/narration, where the concern is how to tell the events
- is how to properly convert the events into forms adequate for dramatization.
