Information Technology Reference
In-Depth Information
event spins into a narrative structure. Rhetorical devices are sentences (actu-
ally declarations of NLG rules and arguments) that can connect episodes or
collections of episodes together to create a story flow. For example, the sentence
“Yet progress doesn't always yield satisfaction” can be used to connect several
episodes describing the positive effects of technological progress and several
episodes describing social or environmental problems arising from technolog-
ical progress. Associated with the English sentence is a formal representation
constraining the meanings that episodes before and after the rhetorical device
can have. For example, “Yet progress doesn't always yield satisfaction” has con-
straints specifying that everything preceding the rhetorical device must be posi-
tive technological, artistic, or industrial progress, and that everything following
therhetoricaldevicemustbenegativeeffectsofprogress.
Once a collection of spins has been connected together by rhetorical de-
vices, the resulting story, which at this point still consists only of formal rep-
resentations, is sent to the natural language generator to produce the actual
narrative text. In addition to generating text, the natural language generator
associates index terms with each generated sentence. These index terms are
used to retrieve appropriate movie and sound clips from a term-indexed mul-
timedia database. Even though the mechanisms linking images to narrative are
less sophisticated than the mechanisms producing the narrative, the Kuleshov
effect ensures that the resulting juxtaposition of image and narrative will still
make sense to the audience.
This architecture was arrived at through an artistic as well as technical ex-
ploration. We desired an architecture that creates narratives rendering our au-
thorial intent without necessarily portraying our own ideological viewpoint.
Through such an architecture we can see stories created that might involve
unusual causal relationships or unexpected conclusions, that, while satisfying
us as authors, go beyond our own conceptions. Additionally, the history con-
struction process captured in the software architecture is itself of conceptual
interest. We see it as a caricature of ideological thought and “cookie-cutter”
documentary construction, an explicit comment on the mechanical nature of
shallow ideological reasoning. Our engagement with AI in the
Te rmi na l Time
project is a concrete example of expressive AI. The AI architecture serves the
needs of, and simultaneously informs, our artistic intent.
