2011/06/16

[Draft] The Spectrum of Stories: decomposing stories with multi-scale sequences

This is a draft of some preliminary ideas and thoughts developed in the past few years.  The concept is not yet complete nor solid, so for people who are interested, please feel free to leave comments.
[Intro]
Mathematic models have been proven useful in many domains, ranging from physics, engineering, biological systems, to human behaviors and finance market.   However, there are still many more fields remain intact from this symbolic and logical way of interpretation.  Some are yet too complicated for current mathematical tools, such as the dynamics of an evolving network, and some are protected by certain long-lived belief, such the human soul. (and I know, there should be many citations.....)

Stories, the term for a recounting of a sequence of events or narratives, play an important role in communications nowadays.  By telling a story, the receivers (listeners) get emotionally aroused and hence are more easily to accept or to understand the messages sent with the story.  Traditionally this field belongs to art and humanity, and it is believed that there must be something unquantifiable here which prevents the invasion of math.

Here we want to propose a mathematical framework for stories, or at least the structures of the stories.  In script-writing, a drama can be divided into episodes, and an episode can be divided into scenes, and hence events.  And what is the composition of an event?  The "Five-Ws" concept from journalism, which is regarded as basics in information-gathering, can serve as a foundation.  Hence, a story can be decomposed into sequences in different scales. Here we use "scale" instead of hierarchy, because a scene can go through two or more episodes, so can an event to scenes.  Since those sub-structures are not necessarily in hierarchy, just like wavelets don't have to be in-phase in a spectrum, the term "scale" might be more appropriate.  After the decomposition, a story can be seen as being "transformed" into a spectral space, and hence we call it a spectrum.

Since these sub-structures of a story are sequences, an intuitive way to model them is using "n-gram".  The chances that this approach might work, should be similar to using n-gram of words to model sentences, or using n-gram of sequence of pitch/duration to model melody: no one can guarantee for now, but it seems promising.

Alright, the draft should stop right here, since the technical details should be in a more formal document instead of random murmurs.  This document should just serve its own purpose.




沒有留言: