21 Dec 2006

At the moment, in my attempt at Interactive Storytelling, I'm thinking about storyworlds. Worlds, in which stories can be told.
My topic of interest is now language.

(Wait,wait,I'll explain the picture later)

Language, normal written language, is the only real material I am working with.
It is the mind and the matter.
In advanced applications, you also want other types of language, like body language or facial expressions. They will make experience complete. But basically, they don't carry the meaning. You can read a book and use your imagination to enrich the meaning with experience that way.
Written language, therefore, is not a problem that I'll handle at the beginning and then can go do other interesting stuff. I will deal with it all the time.
As will become clear later in this text, every single time I think of what the system can or should do, I will return to the language to make it sayable. So, in a way, such a system has some kind of Sapir-Worf hypothesis as inbuilt premise (that's a famous and much disputed linguistic hypothesis saying that you can't think what you can't say).

The point here, and that is what makes this task so interesting to me, is not to come up with a system that can communicate about anything in any style. I now about the infinity and complexity of language. It is out of my respect that I know that the best thing I can do now is a great illusion. My success is measured by the difference between the effort I put in and the experience I actually accomplish.

In a way, you can compare Interactive Storytelling at the current stage to the movies a hundred years ago. The technology was only evolving, but you could come up with ingenuous methods to create an illusion that really entertained people.

For instance, the "Ames-Room" trick to make one person seem ridiculously small compared to others, like Jim Carry in "Eternal sunset of the spotless mind" when he goes through his childhood again in his dreams. In the bonus material on the DVD they explain how they did it. They were not using computer effects, they were doing it the old-fashioned way.
The way the innovators of cinema did it 100 years ago. They build a room that changed perspective by narrowing walls. If you see it from one perspective, people that are standing 10 meters apart actually seem to be next to each other. But of course they are seen in different sizes. This trick is old, but still used.

Maybe some tricks that are invented for Interactive Storytelling these days will still be used 50 years from now.

Ok, back to our actual technical problem: language. Language makes it possible to say anything. We don't want that.
In fact, the truth is, we can't handle that. Because every sentence must be transferred to mean something.
When the user wants to say:
"I kick Fred"*
we must have a Fred represented in our system and calculate his hurt he experienced by the kicking. Earlier, we would have to know that "shoot" is a verb and it goes with one transitive object.
Or maybe we said it can have two:
"I kick Fred in the butt"
So now the first transitive object declares who we kick, and the second declares where we kick him/her.

So that's introducing the structure problem. That's what grammars are for. The trouble is only just beginning.
Can we only kick people? I could kick a dog, too. Or a vase. Do I need an ontology of things in the world, so that I can kick anything that's build of matter? That relates to the problem called "semantics".

Now to the really hard stuff: Meaning. If I kick the vase, what does it mean to Fred? Is it his vase? Does he like it and should now be sad? Am I showing aggressive behavior and he should be fearful? Or are we drunk teenagers and it actually doesn't mean anything but fun (and a little danger of getting caught)?

That's a hell lot of stuff to chew on. Some might say that more computer power and advances in language processing technology that rely on statistical methods on large natural language corpora will someday overcome those problems. Ok, some things will change. Translation, for instance, is on the edge to become really good within a few years. But at least the meaning problem is different. No brute force will solve it.
This is were some really smart engineering should do dome magic. Not to solve it, mind you, but to allow building applications that give you the feeling that they can hadle what you intend to say about the world it deals with.

The trick that I'm after here, that might enable me one day to deliver an illusion that people will bother spending their time with, will have something to do with the creation of a sublanguage.
A sublanguage that allows me to keep control of everything that needs to be done to turn utterances into things happening in the world. To compute reactions. At the same time, using the sublanguage shouldn't make the user feel restricted.

Within the huge space of possibilities that language offers, I need to carve a room that's handy to work in, but also comfortable to live in.
Some rooms can made to seem larger than they are, like they do it in the movies. Sadly, I can not force the user into one and only one viewing position.

* using the name "Fred" is a reminiscence to Chris Crawford's great book "On Interactive Storytelling". He uses this name for examples a lot.
# lastedited 21 Dec 2006
You are seeing a selection of all entries on this page. See all there are.