| Popis: |
We are designing a system to model narrative structures as expressed in lightly formalized text-centric chunks. The system is novel in shifting much of the organizational com- plexity to a categoric reasoning system in a two sorted logic. We optimize for two goals. One goal is to grow a pool of stored insights concerning complex narrative constructions, learned by aggregating crowd-sourced insights. These em- phasize qualities that escape capture by existing methods, and include deliberate ambiguities, poetic allusions, irony, self reference, dynamic reinterpretation and cinematic de- vices. A second goal, described here, is to present suggested, machine generated narrative paths for an unskilled user, gen- erated on the fly and informed by developing insights of mul- tiple narrative situations. The narrative paths may follow the target video's narrative, annotative machine constructed essays or some synthesis of these. |
