Problem of the current storytelling system

I played with the system for a couple of hours using Glorianna's examples. It turned out that the system works pretty bad, if we want it to give a series of clips based on a small paragraph like this,

"We decide to build some sensors for the dance. the choreographer invents a dance. the choreographer teaches the dancers the steps. a dancer had difficulty with one of the movements. she had to practice again and again."


There are many potential reasons, but I think the major reasons include the following:

1. The system works only on the "tokens" For the first sentence, it just collects "decide", "build", "some", "sensors", "dance", and can be regarded similar to any sentence that contains these concepts in a very different order/co-relation.
2. The system doesn't take into account the structure of a sentence. It doesn't recognize who is the main character, what is it that he/she does, who are the other participants in this event, and so on.
   
3. No relationship between sentences is considered. The sentences are processed individually, and the system cannot figure out what the audience would anticipate based on an event in one sentence, and try to find it in the following sentneces. It requires the understanding about relationships like conflict-negotiation in order to do that.

However, if the system is allowed to find several (say, up to 5) recommendations upon each input story sentence, then usually I can find something interesting/related in the results. In other words, if the users can choose from 5 results every time they input one sentence, and this kind of process proceeds iteratively, it might be a more interesting storytelling process, compare to those where only one video clip is found when one sentence is inputted. The reason is that, the activity will feel much like storytelling with recommendations, instead of storytelling with exact search. So, a better way to do it (to make it published) might be changing the interface and the interaction mechanism, and have users to play with it. The next step, namely, taking into account the sentence structure and the narrative, should be a big step and requires a bigger amount of time.