Information Technology Reference
In-Depth Information
In this chapter I have described three of the works that constitute the project on
narrative generation : the individual plot generator, the Evaluator and the collective
plot generator. Together, they combine more than 15 years of research. The way the
project has progressed shows some of the diverse courses that research on narrative
generation can follow. It also illustrates the value of building new models on top of
previous works.
This chapter can be summarized as follows. The ER Computer-Model provides
plausible explanations about how important aspects of the creative process of writing
operate. It represents the creative process as a continuous cycle between automatic
generation of material during engagement and the evaluation, and if it is necessary
the modification, of such material during reflection. The evaluation process produces
a set of guidelines that constrains the production of material during engagement. In
this way, stories emerge as a result of the ER-cycle; that is, there are no predefined
story structures or explicit character goals. The system first generates sequences of
actions and then goes back to check that all actions' preconditions are satisfied. These
characteristics distinguish MEXICA from other computer models of writing.
An extra singularity is the organization of memory in Contextual-Structures; such
structures are comprised by emotional links and tensions between characters. As
far as I know, there is no other model that employs emotional links and tensions
between characters to represent the state of affairs of a narrative. It turns out that this
representation provides great flexibility during plot generation. These results make
me wonder about the essential role of emotions to make sense of the world and on
their utility to deal with endless circumstances through life.
There are “external” components of computer models that are hardly mentioned
but, nevertheless, play a significant role in it. They are components that “we take
for granted”, so to speak. MEXICA offers an opportunity to reflect on them. An
example is the information that the agent exploits to construct its knowledge base
and which is supplied by the user: the pre and post conditions, the story actions and
the previous stories. Based on the ideas expressed in this text, it would be useful
to work on developmental agents that allow observing how emotional links and
tensions emerge during development, how they are associated to particular actions,
and how the cognitive abilities necessary to represent them appear. Some efforts in
that direction have been done (e.g. see Aguilar and Pérez y Pérez [ 18 ]).
Evaluation is a recurrent topic in meetings about computational creativity. Despite
its relevance we hardly understand how it works. Human evaluation is too complex;
it depends on individual knowledge and experience, cultural and social traditions,
and even personal mood. That is why it is so hard to model it. Although some authors
have contributed in this area (e.g. [ 19 - 22 ]) much more work is desirable. The main
goal of the Evaluator is to help to understand better this key process. MEXICA and
MEXICA-impro provide a useful framework for this task.
The model for the generation of collaborative narratives includes a cognitive
dimension and a social dimension. Because each agent has its own knowledge base
and the capacity of producing interesting plots, their interaction allows exploring
the links between cognitive and social aspects during plot generation. We need
Search WWH ::




Custom Search